UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

A television camera to computer interface Yusuf, Tundey 1970-05-24

You don't seem to have a PDF reader installed, try download the pdf

Item Metadata


831-UBC_1970_A7 Y68_2.pdf [ 3.47MB ]
JSON: 831-1.0102116.json
JSON-LD: 831-1.0102116-ld.json
RDF/XML (Pretty): 831-1.0102116-rdf.xml
RDF/JSON: 831-1.0102116-rdf.json
Turtle: 831-1.0102116-turtle.txt
N-Triples: 831-1.0102116-rdf-ntriples.txt
Original Record: 831-1.0102116-source.json
Full Text

Full Text

A TELEVISION CAMERA TO COMPUTER INTERFACE by TUNDEY YUSUF B.Sc. University of Wales, Swansea, 1968 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF APPLIED SCIENCE in the Department of Electrical Engineering We accept this thesis as conforming to the required s tandard Research Supervisor Members of the Committee Acting Head of the Department Members of the Department of Electrical Engineering THE UNIVERSITY OF BRITISH COLUMBIA August, 1970 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the Head of my Department or by his representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Depa rtment The University of British Columbia Vancouver 8, Canada ABSTRACT . This is an instrumentation thesis. The interface system discussed is a link between an ordinary TV camera and a computer for storage of visual data. The same system can also be used as a link between the computer and a display monitor. Because of its wide bandwidth, a video signal cannot be sampled at the Nyquist rate and presented to a computer. Previous interface systems overcame the problem by scanning slowly on an .element-by-element basis using a special scanner and then presenting the samples to the computer. After processing, the data would be read out at the same slow rate and displayed on a special display monitor. The interface described in this thesis will accept material obtained from an ordinary TV camera scanning at standard rate. By using a "stroboscope" sampling technique the samples are pre sented to the computer slowly enough for it to process. After processing, the data is displayed in a similar manner on a normally scanned monitor for evaluation. Basically the interface operates as follows: A TV camera vid-ao signal is sampled at a rate slow enough for i computer acceptance. The camera scans the same picture several hundred times until all the points representing the picture have been sampled and stored, the sampling is controlled such that all the points are each sampled only once. Because of the sampling method consecutive samples in the computer do not correspond to adjacent points on the picture being stored. It may therefore be necessary to programme the computer to arrange the samples such that adjacent data in the computer represent consecutive picture points before processing. After processing, the samples may be rearranged and read out for display in the same order they were stored. ii The horizontal resolution of the picture being stored can be varied quite easily in steps. For example, a system designed to have a maximum of 480 points/line will also have the ability to provide such lower resolutions as 60, 120 and 240 points/lines. This variation is made possible by the design of the hardware. By software the vertical resolution can be varied between an upper limit of 525 lines per picture and such near submultiples of this as 263 and 131 lines/picture. The thesis is discussed in relation to the PDP-9 computer on which most of the work described was done. However, the system interfaces readily with other computers. iii TABLE OF CONTENTS Page ABSTRACT ii TABLE OF CONTENTS v LIST OF ILLUSTRATIONS v 1. INTRODUCTION 1 1.1 Previous Interface Systems 1 1.2 The New Interface System 2 2. HARDWARE REALIZATION . . 7 2.1 The Storage System2.2 The Display System 15 3. PROGRAMMING TECHNIQUES . 23 3.1 Coarse Quantization 24 3.2 Storage Programme for 7-bit Quantization 26 3.3 The Display Prograjnme 28 4. PERFORMANCE TESTS 31 4.1 Qualitative Tests4.2 Quantitative Tests 5 5. SUMMARY AND CONCLUSIONS 36 5.1 Limitations5.2 Methods of Improvements 37 5.3 Possible Uses 40 APPENDICES 41 BIBLIOGRAPHY 7 iv LIST OF ILLUSTRATIONS Page 1.1 A Typical Visual Data-Computer Interface ..... 3 1.2 The New System Interface « 3 1.3a Picture being Scanned 5 b Video Signal Showing Points where Samples Have been Taken • 5 2.1 The Storage System 8 2.2 The Frequency Multiplier 10 2.3a,b,c,d Multiplier Waveforms 1 2.4 Pulse Amplifier . 12 2.5a,b Sample and Hold Circuits 13 2.6a Sample and Hold Input 4 b Sample and Hold Output 12.7 The Display System . 16 2.8 Computer D/A Output 8 2.9 Multivibrator Delay 19 2.10 Pulse Amplifier 20 2.11 Gated Operational Amplifier 21 2.12 Sawtooth Generators 2 3.1 Flow Chart for Programme to Store and Display a Picture at. 4 bits/sample 25 3.2 Flow Chart of Programme to Store a 7-bit Quantization Picture on PDP-9 Tapes 7 3.3 Flow Chart of Programme to Continuously Display.the 7-bit Picture 29 v Page 4.1a Original "Test Pattern" Picture 32 b Output from Computer at 138 usec/sample 3c Output from Computer at 69 ysec/sample 2 5.1 Phase Locked Loop Scheme 38 5.2 * Generation of a Narrow Pulsevi ACKNOWLEDGEMENT I wish, to express my gratitude to the National Research Council of Canada for the financial assistance in carrying out the research and to Dr. Micheal Beddoes for very useful ideas and advice. I would also like to thank Dr. J.S. MacDonald for reading the thesis and making useful suggestions and to Rodney. George, Joan Kushner and Heather DuBois for the preparation and typing of the manuscript. vii 1 1. INTRODUCTION The use of computers to process visual data has been in existence for some time. The first problem that researchers in the field had to solve was how to get the visual data information into the computer. This was done by scanning the picture on a point-by-point basis in a TV camera-like manner to obtain an electrical signal equivalent of the visual information. An ordinary TV camera was not used to act as the link between the picture and the computer because of the wide bandwidth of the resulting video signal. This would have caused a data presentation rate greater than the computer could accept. Instead special cameras and display systems were designed to interface the visual information to the computer. This thesis presents the results of an investigation into the possibility of using an ordinary TV camera and display monitor in visual data processing. As a prelude to discussing this and some of the earlier interfaces, a brief description of the way a TV picture is obtained will first be presented. A TV picture is made up of lines which are sequentially scanned and presented to the viewer. Standard TV presents one line of picture in about 67 usees. A total of 525 lines are used to complete a picture. Just as each picture is made up of lines, each line is made up of points. The number of points per line is determined by the resolution of the camera. This resolution varies from about 100 to 650 points per line. Thus a TV picture is made up of points and any system required to store the picture information will only need to have the ability to store the information of the points. 1.1 Previous Interface Systems In previous picture-computer interface systems, the points making up a picture were scanned sequentially. The rate of scanning was slowed down 2 so that the computer could accept the information. A typical system is shown in Fig. 1.1 and operates as follows: A transparency of the picture is scanned by a slow moving flying spot scanner. The resultant electrical signals, representing the brightness of successive picture elements are A/D converted and recorded on magnetic tape. This provides a digital version of the input picture for computer processing. After the picture has been processed according to some algorithms, the computer instructs a microfilm printer to generate the new product. This is done in a manner which is the inverse of the original scanning. A TV-like tube paints the picture with a moving spot of light and a camera records it on film to provide a photographic end product There are variations of this basic method. One such variation eliminates the need for tape recording by inputting the digital data straight into the computer memory. In such a case the flying spot scanner may be required to wait at each picture point so as to give the computer sufficient time to accept the data. Either way the system is complex and costly. 1.2 The New Interface System We are here interested in being able to use an ordinary TV camera, scanning at standard speed. Obviously the points on the picture cannot be sampled sequentially since this rate of presentation is too fast for any computer. The problem is overcome this way. A point at the beginning of the picture, say, is sampled and stored in the computer. Assuming it takes N Liseconds to store this and a resolution of 120 points/line is required, at the scanning rate of 67 useconds per line, the time between successive picture points is ^ useconds. Therefore ^57^ points away from the first point will correspond to a time of N useconds. This is sampled as the next point for the computer. Another In PUT TRANSPARENCY FLYING--SPOT \ • A/r> CONVERTER DIGITAL TAPE RECOVER COMPUTE^ TAPE RECORDER PICTURE PHOTO&MPHIC EHLfjRGER, MitROFlLN] PRINTER Fig. 1.1 A typical Visual Data-Computer Interface, OPVINAR. y PICTURE OR ST/^T/ONSRY OBTEC-T V SYSTEM INTERFACE CofAPOTELR i T>iSPLfiy FROCESSEI) MaNITOR^ gs- PHOTOGRAPH Fig. 1.2 The New System Interface. 4 points away, a sample is. again taken and the value stored. The process 67 is repeated until the bottom of the picture is reached. When the scanner , ., t . t • 120N . . returns to the top, a sample is taken at a poxnt that is ^ points away from the last one taken at the bottom of the picture. As can be seen from Figure 1.3, consecutively numbered points, which represent adjacently stored samples in the computer, are not adjacent picture points. Therefore, the camera must scan the same picture repeatedly several hundred times until all the points making up the picture have each been sampled and stored. The sampling is controlled such that all the points are each sampled only once. It can be easily shown by induction that this can be done if a number p is chosen as the number of picture points between successive samples where p is not a factor of and does not have a common factor with the total number of picture points, p must also , • 120N be greater than . 6 / The interface stores the data in the computer under programme control. When all the samples representing the picture have been stored, the computer may have to arrange them so that adjacent samples in storage are adjacent picture points before the.data, can be processed as desired. When the data has been processed and is ready to be displayed, a reverse process to the one described above now takes place. The samples must be in the same positions they were when first brought into the computer. The data is then read out in the same way it was stored. The first information out of the computer goes to the beginning of the screen display. The second sample is actually ^"32^ points away, so the screen must be blanked until b I this point is reached. When the point is reached, the screen is unblanked and the point displayed. The next sample is displayed again "^y^" points away from the last one. The process goes on until the bottom of the screen is LINE 1 ___ LINE 2 LINE 3 LIME -4 /./NE 525 Fig.1.3a Picture being scanned (0 07) SAMPLING-PULSES Fig. 1.3fc.Video signal showing points where samples have been taken. 6 reached. The scanner returns to the top and displays the next sample at points away from the last one at the bottom of the screen. These steps are repeated until all the points have been displayed. If the time between samples is 69 usee, a total time of about 3.8 seconds would be required to build a picture represented by 63,000 points. This is the same time it takes to store the picture. An ordinary photographic camera focused on the display screen and opened for 3.8 seconds will give a photograph of the processed visual information. A block diagram of the interface system is shown in Figure 1.2. Because of its dual role the system has two parts: a storage part and a display part. The storage part consists essentially of a sample and hold circuit and control unit. The control unit decides the time at which sampling of the visual information is to start and the time between samples. It is made up of a frequency multiplier, counter and pulse shaping digital circuits. The display part consists of a set of monostable multivibrators, sawtooth generators and a gated amplifier in addition to the control unit mentioned earlier. This interface system has several advantages over some of the existing ones. Among these are the ability to use an ordinary photograph or even a still object as the visual information source instead of a transparency, the possiblity of working with existing ordinary TV camera and display equip ment, the ease of varying the picture resolution and the low cost of the hardware. The limitations of the system and methods for improving its per formance will be discussed in Chapter 5. The next Chapter will be concerned with a detailed description of the hardware while Chapter 3 will give a review of the programming techniques required to operate the interface. Results of performance tests on the system will be presented in Chapter 4. / 2. HARDWARE REALISATION 2.1 The Storage System The scheme presented here requires the knowledge of the exact number of points making up a picture. Since the manufacturers only give an approxi mate value with their cameras, it is necessary to find a way of obtaining an exact value. It is also necessary to count the points so as to know when to take a sample during storage and when to unblank during display. In order to know the exact number of points making up a picture, it would have to be made up by the designer. The 67 useconds line scanning rate of standard TV camera corresponds to a line frequency of 15.5 Khz in a 525 line per frame system. There is an oscillator in the camera going at this frequency and causing the line scan. If this frequency is multiplied by 120, say, an effective line resolution of 120 points is automatically given to the picture under observation; each cycle corresponding to a point. Actually a TV camera has a master oscillator going at twice the line frequency. In the system described here, a "multiply by sixty" circuit was used to get the desired 120 points per line. The multiplier output was used as a clock to a counter, each clock cycle corresponding to a picture point. The counter had a module p-1, so each output of it corresponded to a point to be sampled on the video signal. Since the camera will scan the same picture repeatedly, it is only necessary to keep a count of the number of samples already taken to know when sufficient data has been stored to represent the picture completely. This was done by software. The camera also provides a synchronising pulse each time the scanner goes back to the top of the picture . If the computer waits for one of these before starting to store, the first sample taken will be at the beginning of the picture. r MULTIVIBRATOR ' 1 I START OF I S7-otf<?6£ MULTIPLIER ZN-TE&FftCE SYSTEM Fig. 2..2 The Storage System. 9 A sample and hold circuit was used to hold a sample long enough for the computer to convert the analog signal into a digital value. Actually each sample was held until the next one was taken. The computer will thus have plenty of time to A/D convert and store each sample. The storage part of the interface is as shown in Fig. 2.1. Each of the circuits making up the system will now be described briefly, beginning with the multiplier the diagram of which is shown in Fig. 2.2. The basic principle behind the multiplier is that of non-linear ampli fication and filtering. The output of the camera master oscillator is shown in Fig. 2.3a. This is very rich in harmonics. It was passed through a resonant circuit of centre frequency 5f where f is the fundamental. The output of o o the resonant circuit was used to drive a CE amplifier and because of its large amplitude, it caused non-linear amplification. The output of the amplifier is shown in Fig. 2.3b. A resonant circuit tuned to f^=^f^ where f^=5fQ was used to pick the fourth harmonic. Thus a multiplication by 20 was obtained. The new frequency was again non-linearly amplified to produce another harmonic-rich waveform. A further multiplication by 3 was obtained by tuning. Since f=2x line frequency, the final output had a frequency of 120x line frequency. In other words, a system with 120/line resolution was produced. If a better resolution was required, it could have been easily achieved by further multipli cation and filtering. This method of multiplication thus allows for a wide range in the choice of the horizontal resolution. It is interesting to note that induc tors were used in the resonant circuit instead of transformers. As long as their Q's are good enough, inductors work satisfactorily in this type of circuit. The use of emitter followers between stages is also to be noted. The output of the multiplier was used as a clock for the counter. Fig. 2.2 The Frequency Multiplier FIG. 2.3a TV-Camera master oscillator output f » 31KH . 20 usec/division FIG.2.3c Multiplier second stage output f = 4x5x31KH = 620KH . z z 5 ysec/division FIG. 2.3b Multiplier first stage output f - 5x31KH - 155KH . z z 10 usec/division FIG. 2.3d Multiplier third stage output f - 3x4x5x31KH = 1.86MH . z z 1/2 ysec/division -9 v cc 5£k •-lOv OUT Fig. Z.4 Pulse Amplifier The counter used in the system is essentially the parallel type described in the literature. After amplification by the CE circuit shown in Fig. 2.4, the counter decoder output was used to sample the video signal. In order to sample the video signal at the desired time only, a circuit was required which, blocks the signal from going through at any time other than when a sampling pulse is received from the counter. Furthermore? the sample taken at that time must be held until the next sampling pulse so that the computer has sufficient time to A/D convert and store the value. The sampling window must not be more than about 300ns wide if the desired horizontal resolution is to be achieved. A sample and hold circuit was re quired to the above specifications. The basic circuit for a sample and hold is shown in Fig. 2.5a. The slew rate of the required operational amplifier is extremely high and could not be met by those available at the time the interface was being designed. Alternative methods of sample and hold were investigated. It was found that a sampling scope has a sample and hold in it with the following characteristics: sampling window of 50 nseconds, J. J Fig. 2.5a Simple Sample WHEH THB SWITCH IS WHEN THE SWITCH ;S and Hold Circuit. CtOSEb VOUT ^ V>« SWITCH SWITCH MEMORY PUL$t5 Fig. 2.5b Sampling Scope Sample and Hold block diagram. VJHEN THE SWITCH »S CtOSEi V0yT = V,H 14 FIG. 2.6a TV Camera video signal (input to sample & hold) FIG. 2.6b Sample and Hold output hold time up to 500 usee with negligible decay. It was designed using valve amplifiers. To save time, a similar design was not built, instead a sampling scope "plug-in" unit was used as the sample and hold in the interface system. The block diagram of the unit is shown in Fig. 2.5b. Normally the scope generates its own sampling pulses but in this system it was provided with sampling pulses by the counter. The system worked well as can be seen from the pictures of Fig. 2.6 which shows the video input and the sample and hold output. The signal is now ready for input into the computer. 2.2 The Display System When the visual data has been processed according to the desired algorithms, it is required to be read out of the computer and displayed for evaluation. The system interface will act as a link between the computer and a display monitor. Its operation in this mode is the reverse of that described in the previous section. The block diagram for the display part of the interface is shown in Fig. 2.7. The multiplier and counter from the storage system are also part of this. In addition, there are 2 sawtooth generators, a gated operational amplifier and some multivibrators. The counter output pulses are now used to instruct the computer to display the corresponding point on a monitor instead of storing. We wish to display the points on the monitor in the same order that they were stored. The digital bits in the computer, representing the samples, were first D/A converted. Each D/A output was an analog signal lasting until the next sample. The waveform shown in Fig. 2.8 was obtained from the com puter. This is similar to the output of the sample and hold, Fig. 2.6b. In order to get point display for each analog level, the levels were sampled with r ~ FRAME PULSES o MyLTivieR/)To« MULTIPLIER COUNTER. 1 PULSE MULTIVIgfi/^ PULSE r r — 4 LINE puisrs. FRAME ftVLTlVIBRfiTdk PBLfiy SfiVJTooTH GENERATOR X I PROCESSED I V _1 si eiv/rtL. C R T L J Fig 2.7 The Display System a narrow pulse by passing them through the gated operational amplifier. The gate of the amplifier was controlled by the sampling pulses from the counter. The short duration analog signals obtained were then used as the "video" input to the CRT. Since the signals had no synchronizing pulses, unlike a normal composite video signal from a TV camera, external pulses were provided to cause horizontal and vertical deflections. These pulses were obtained direct from the camera. A TV monitor was not available during the design and building of the interface. The system was therefore designed for display on an ordinary oscilloscope. The amplifiers were used as "plug-ins" to the scope and each amplifier had a sawtooth waveform applied to it. One sawtooth was generated from the horizontal camera deflection pulses and the other from the vertical pulses. There is usually a few microseconds delay between the time a computer receives the pulse instructing it to output a stored sample and the time the sample is actually received. Since the instructing pulse is also used to control the gate of the gated operational amplifier to allow a converted signal through, it was delayed by the few microseconds before being applied to the gate so as to give the computer time to output the value. The vertical and horizontal pulses were delayed by the same amount before being applied to the sawtooth generators so as to preserve the relative positions of the points on the screen. , The exact value of the computer delay is not known but is about 5 useconds for a PDP-9. A delay of 10 usee was given to each pulse. This gave the computer plenty of time to convert a value and settle down. The delays were provided by monostable multivibrator circuits. These circuits also acted as shapers of the various pulses. Fig. 2.8 Computer D/A Output Fig. 2.9 Multivibrator Delay £80 / ^-WA IN our 0 OUT Fig. 2.10 Pulse Amplfier The sawtooth waveforms applied to the scope caused a raster to appear on the screen. The output of the gated operational amplifier was applied to the CRT blanking and thus z-modulated the screen. The intensity due to the signal was superimposed on the raster. The intensity was turned down until the raster disappeared. The only things then visible were the random points that add up to make the picture. By taking a time-exposure photograph of the points, a picture was obtained. Most of the components making up this display part of the interface system are off-the-shelf I-C modules. The multivibrator delays were made from Motorola MC 799 dualvbuffer gates. The circuit for one such delay is shown in Fig. 2.9. The only external components are resistors and capacitors. The output of the first multivibrator was differentiated. The positive half was then used as input to the second multivibrator. The width of the first pulse thus determined the amount of delay applied to the input pulse. A Motorola MC 1545 gated operational amplifier was used to obtain the narrow analog signals which z-modulated the scope. The amplifier works as follows. A signal applied to its input appears at the output if the gate 4.7 K OUT + 5' Fig. - £••// Gated operational amplifier has 5v or more applied to it. If the gate voltage is less than this the input is inhibited. In our system the gate was controlled by the delayed counter output pulses. The pulses were first differentiated and then amplified by the common emitter circuit shown in Fig. 2.10 before being applied to the gate. The differentiation was necessary in order to obtain the required narrow analog samples. The circuit for the sawtooth generators are shown in Fig. 2.12. They are standard sawtooth circuits and therefore need no description. The input and output systems have now been described. In the next chapter the software that controls the operation of the system by the computer will be discussed. 560-IN 0,1'fuF 2Z0 MPS 637 .00/ 4 70ff -h25v OV7 Fig. Z-f2.°-Hor izontal Sawtooth Generator OOJ 1— —i i—. %— Fig.z./z'b Vertical Sawtooth Generator 3. PROGRAMMING TECHNIQUES The philosophy behind the programme that operates with the computer will be presented in this chapter. The systems interfaces readily with other computers but the actual programme presented here are for the PDP-9 machine on which most of the work was done. Any computer that is to be linked to a TV camera through the inter face system must have A/D and D/A conversion capability, a large core memory or a fast mass storage facility. The computer must also be able to sense when to start storing the visual data and at what rate. Since the sampling speed of the interface is geared to the rate of input of data into the computer, a large core memory or a disc in the computer would be ideal. When the signal is given by the interface (see Fig. 2.1),' the computer starts to sample the visual data. Every sample is A/D converted and stored in the memory. Each time a sample is stored, a count is taken and when this count reaches the number of required samples, sampling ceases. Since consecutive samples stored in the computer are not consecutive elements on the picture being stored, the computer may have to arrange the points so that adjacent samples are adjacent points of the picture. The data can then be processed to simulate a desired system (a separate programme is required to do this). After processing, the samples must be in the same relative positions they were when first brought into the computer. The computer then performs a D/A conversion on the samples and outputs them for display through the interface system. The computer is programmed to start storage or display only when the camera is at the top of the picture scan. With a large core memory virtually any computer can be programmed to carry out the steps described above. A PDP-9 for example can have a core memory of 32,000 words, each word 18 bits long. If each picture sample is converted to 7 bits, say, five samples can be packed in 2 computer words. Thus only 25k words of memory would be required to store a complete picture of 525 x 120 i.e. 63,000 points. From the memory the data can be stored on an auxilliary storage facility for later processing. The typical time for A/D conversion and storage of a sample is 20 usee, therefore total time to store a picture would be about 1.3 seconds. With a faster A/D, this time would be reduced. 3.1 Coarse Quantization More often than not, small and medium scale computers do not have a large core memory. The PDP-9 used, for example had only 16k of memory. With coarse quantization (4 bits/sample), it was possible to store the information of a picture in this memory. Nine samples were packed in two computer words. Thus only 10.5k words of the memory were needed for the 63,000 samples. The optimum sampling speed of 50Khz could have been achieved. However the interface system was designed for only two sampling speeds: approximately 7.25 Khz and 14.5 Khz corresponding to 253 and 121 points respect ively between samples. In the case of 7.25 Khz sampling rate, about 3.8 seconds are required to store a complete picture as opposed to twice that for 14.5 Khz. The programme flow charted in Fig. 3.1 (actual programme is included in the appendix) was used to store the picture information in the computer. When sufficient samples have been stored in the memory, the data was transferred to a tape. The computer then went to a subprogramme which arranged the samples as described previously and processed them as follows: Each sample was com pared with the one before it. If the first three bits were equal, the samples were taken as being of equal amplitude. If the three bits were unequal, the value of the present sample was retained and compared with the next one. SET UP TAPE TO ITS BEGINNING-f Wfi/T FOrX TOP OF F/CTd*£ POLSE FO/Q, SfiMPLtNG-PUL&E SbmpLE ^ ft /j> CONVERT SHIFT <S- STORE DEPOSIT ON TAPE. ! RETOKN fR0V\ SUBPROGRAM * W/9/T FaP. ToP OF PICTURE PULSE J^ESBT REGISTERS UNPACK j v/ft CONVERT dvTPUT &• SlSPlfly I tV/9/7 FOR, sfimPUNS-PULSE Fig. 3 i Flow chart for program to store and display a picture, h bits/sample. All the samples were processed in this manner. This simple process is in fact a simulation of a possible digital transmission scheme whereby during any par ticular sampling period, data is sent across the communication channel only if the present sample exceeds the last by a predetermined threshold. After processing the samples were rearranged, in preparation for read out and display in the same order that they were stored. The computer returned to the programme of Fig. 3.1 and started displaying the processed data. The samples were read out of memory repeatedly so as to provide a continuous display of the picture for as long as was required. The computer waited for a "top of picture" pulse before starting to output and display. This ensured the synchronisation of the "video" signal with the horizontal and vertical deflections on the screen during the display. 3.2 Storage Programme for 7-bit Quantization With a disc mass storage facility, no difficulty would arise in using a 16k system to obtain 7 bits/sample or higher resolution. A typical PDP-9 disc takes about 1& usee to transfer a word of data to or from memory. It would therefore be possible to operate at the optimum sampling rate of 50 kHz. The PDP-9 used did not have a disc, instead it had a magnetic tape as its mass storage facility. The tape has a transfer rate of 200 ysec per word (+ 20% depending on the diameter). By packing 5 samples in 2 computer words it was possible to store a complete picture information on the tape through the memory. The computer can transfer data from part of its memory on to the tape at the same time as some other data is being deposited on another part of the memory. At a sampling rate of 7.25 kHz, the time required to fill up 2 words is about 690 ysec. Since only 400 ysec are required to transfer 2 words of data to the tape, it is possible to overlap the sampling ENTER I MOVE BOTH TAPES To THE BFG-/MNING-f SET Vp COUNT REGISTERS WAIT FOR TOP OF PICTURE SCfthJ Fig 3»2 Flow chart of program to store a 7-bit quantized picture on PDP-9 tapes. and transfer processes. In fact the 7.25 kHz sampling rate was chosen during the design of the interface in order to be able to do this. Figure 3.2 shows a flow diagram of the programme used to store the 63,000 points picture visual information. Each sample was quantized to 7 bits. Five picture samples were packed in 2 computer words. When 8k of memory had been filled, a tape was set in motion to start emptying the memory. Four and half thousand more words of memory were meanwhile being filled up. By the time this was done, 6k of the memory had been emptied by the tape. The computer went back and started filling up the memory again. The tape finished emptying the first batch before the memory was filled up a second time. After the second memory filling, the batch of data was written on a second tape. Thus 25k words representing the picture were stored on the tapes. If necessary a separate programme could read the data from the tapes to memory, arrange them as mentioned earlier and rewrite them on the tapes for later processing. 3.3 The Display Programme The programme shown in Fig. 3.3 was used to provide a display of the picture information stored by the storage programme which deposited,the 25k words on two tapes. The display programme alternately read the two stapes into the memory for continuous display. The first tape was read into the memory. The samples in each word were unpacked and sequentially read out and D/A converted when the signal was given by the interface. When 8k words had been emptied, the first tape was set moving back to its beginning while more data was read from the second into the memory. By the time all the 12.5k o words had been read from tape 2, the first tape was at its beginning and ready to be read again. When 8k words had been emptied from the memory, tape 2 was set in backward motion and more data read from tape 1. Thus by alter-INTE&KUPT SERVICE ROUTINE ENTER ' I  MOVE TQPES TO y . ser UP COUNT START KEADIH& 12.5 K woRis OF DATA /=KOM T4f>E 1 T» IH7EZRUPT OCCURS IAJHEM /^EAi/M(^ COAlPAETEt) Wr»T FoR. Top of-' PICTURE SCAN UWPflCK SAMPLES, F*OM CsmPPTS-% WORI>S. D/A CON v EAT A.wt> OUTPUT ro INTERFACE SE"F T/)f£ 1 MoViWG- S/)CK . ST/}/?T TAPE Z . THTZRROPT M/IIN p«as«/lM ENTER. STOP T/=)P£" MOTIONS. SET U^3 TO START REPi2>ltJG- F&OtA ONE TAPE AND OTHER TAPE B/}CK WHEW IH$TKUCTEI> By MAIN PROG-RAM I RETURN TO M/lfA/ pftoCxRAitA r>C-TUR.E HAS BEEN 2\S.f>lfWEb ONCE. ReTUft.N TO TOP To START VISPLfiy 1N G- THE" WfW 12-5 K 8/rrcH OF 2>AT* '4.5 K. 'AVPITIOHPIL DlSPLfiy^i P UWPACK. MORE WOR-DS ^OM MEMofty ^ 2/SP2/iy THE POINTS Fig 3-3 Flow chart of program to continuously display a picture. nately reading from the tapes the visual data was continuously available for display. Similar programmes to the: ones described here can be written when other types of computers are involved. However great care is required in programming especially when tapes are used for storage, otherwise synchroni sation between the computer data and the camera deflection pulses may be lost for much of the time. 4. PERFORMANCE TESTS Qualitative and quantitative tests were performed on the interface system to check its performance as a useful camera-computer link. The results of these tests are presented here. 4.1 Qualitative Tests In the qualitative tests, a TV camera was focused on an object and the output of the camera displayed on an oscilloscope. A photograph of the display is shown in Fig. 4.1a (Because of the limitations imposed by the scope, a very simple picture pattern consisting of letters written on a white background was used as the object ). The video signal from the TV camera was stored in the computer using the interface system and the programme of Fig. 3.1. At a sampling rate of 1 sample about every 138 psec, (about 7.25 kHz), a total time of about 7.6 seconds was required to store the picture. The data in the computer was arranged and processed by the subprogramme mentioned in section 3.1. A second subprogramme rearranged the samples after processing then read them out for display as previously described. A time exposure photograph of the display was taken by opening the polaroid camera shutter for 7.6 seconds. Fig. 4.1b shows the result obtained. The relative positions of the points making up the picture were maintained despite the "stroboscope" sampling method. If this was not true the printed words would not have come out as they did. The validity of the method is therefore established. The sampling speed was increased to 1 sample every 69 usee. A total of about 3.8 seconds was required to store or display the picture pattern. A photograph of the scope display is shown in Fig. 4.1c. A diagonal scanning pattern is seen to be superimposed on the picture. This arises as 33 follows. The stroboscopic sampling scheme results in a diagonal pattern during sampling and display (see Fig. 1.3). For fast sampling the time to display a diagonal is very short. Thus if the polaroid camera used to take a picture of the points is opened by a few milliseconds less than the exact time required, some points would be missed. This would give rise to the diagonal pattern seen in Fig. 4.1c. For slow sampling the time to display each diagonal is long so inadvertantly shorten the exposure time by a few milliseconds would not impose a noticeable pattern. The quality of the pictures is due entirely to the scope on which the displays were done. This assertion is borne out by the poor quality of the original pictures (Fig. 4.1a). Non-uniform illumination of the screen, the very low resolution of the scope and the non-linearity of the z-axis modulation are largely responsible for this. Four bit quantization was used in the display of the processed signals. For the input picture patterns used, this quantization level was quite adequate. Any higher resolution would not be too meaningful on the scope display. However it is possible to quantize up to seven bits using the programmes described in sections 3.2 and 3.3. The output from the computer D/A converter for such quantization was passed through the interface and the corresponding picture displayed on the scope. While the signal itself looked like the sample and hold output presented to the computer, displaying the picture on the scope presented some problem. For much of the time, there was loss of synchronisation, as a result the picture was split up, part.ofi. it appearing on one side of the scope and the rest on the other side. This loss of synchronisation was due to the few milliseconds lost by the computer iri . starting and turning the tapes around. The problem was solved by making the computer check for "top of picture" pulse after each complete picture display. If synchronization was .lost, it would be re-established before the display began 'again. 4.2 Quantitive Test The interface was tested for noise and jitter. The most likely source of noise in the system would be the sample and hold. The unit is supposed to sample the input video signal and hold the analog signal until the next sample. If another sample is taken before the output amplifier settles down, jittering will result. This jitter effect was tested for in the following manner. The camera was focused on a white sheet of paper and the signal due to this stored on tape through the memory using the program of Fig. 3.2. The 63,000 7-bit samples were then compared (by the computer) with each other to see how many samples have different values from the rest. It was found that only 100 samples were different. Thirty nine of these samples were out by 2 bits, the rest by only 1 bit. 5. SUMMARY AND CONCLUSIONS A system has been presented which can be used as a link in Picture Processing by Computer. The interface makes it possible to use ordinary TV camera and display monitor in such processes instead of building special ones. Results were presented to prove the validity of the method. The advantage of the system over previous interfaces were also presented. The limitations of the interface will be discussed here as well as methods of improving the system to improve its reliability. Some of the uses that the interface may be put to will also be briefly mentioned. 5.1 Limitations Because of the arrangement that may be necessary before a visual data stored in the computer using this interface can be processed, it is not possible to use the interface as a link for real time on-line computer picture processing. For the same reason it is not possible to process a portion of the picture and display it independently of the rest of the picture. If a portion of the visual data is processed, the whole picture will have to be displayed in order to be able to observe the effect of the processing on the portion of interest. For some operations these limitations may make the inter face system unacceptable. An example of this is contour tracing in pattern recognition computer processing. Where this limitation is not important, there is another problem that may arise. This is the fact that a continuous display of the processed data for a long period of time may not be possible on computers with small or slow storage facility. This is because of the loss of synchronisation mentioned in Section 4.1. For example the tape turn around time on a PDP-12 is so long as to preclude the two tape storage and display method described in Chapter 3. The best that can be done on a PDP-12 that has only magnetic tape for mass storage is to store the visual infor mation several times on the tape. The tape has a-capacity of about 1.6 million bits. This can store four 63,000 sample picture information if each sample is quantized to 6 bits. By storing the same picture on the tape four times, it will be possible to get the processed information displayed four times continuously. After that^ the tape has to rewind back to its beginning (about 30 seconds are required for this) before the display can be restarted. This may not be a serious limitation if all that is required is for the display to last long enough for a photograph to be taken. Another problem that may arise with the interface system is the limit of variation of the vertical resolution. The 525 lines/picture of the TV camera is the maximum resolution possible. This should not be a serious drawback however since for most applications this resolution is quite ade quate. The lower resolutions easily obtained are limited to such near submultiples of 525 as 263 and 131. These are obtained by using the interlace facility in the camera. For instance, to obtain 263 lines/picture, the computer is programmed to store the samples from every other field only. These then are the limitations of this interface system, in spite of them it should still be useful as a visual data-computer link. 5.2 Methods of Improvement The philosophy behind the design of the interface is that ,the sampling pulses be locked (ie well synchronised) with the deflection pulses generated by the camera. If this is not so then it is quite possible that the samples taken during the storage of the visual data will not be displayed in their relative positions after processing. In addition some samples may be missed altogether. The use of the multiplier-divider scheme is to avoid this problem. If the multiplier is well designed, the problem is not serious. However even a multiplier of the best design may sometime burst into oscillation independently of the camera master oscillator and synchronisation may be lost for a time. An alternative and more stable way to lock the sampling pulses to the deflection pulses would be to use the phase-locked loop method. In this method, a waveform is generated by a crystal oscillator at the required high frequency, e.g. 1.85 MHz of the prototype. This frequency is then divided down by digital counters (a very much simpler operation than multiplying) to the deflection frequency. The camera deflection frequency is compared with the externally generated one and the error signal is used to control the frequency of the crystal oscillator in a feedback control loop method. Fig. 5.1 shows this diagrammatically. If either the camera or the oscillator drifts from the prescribed frequency, the other follows automatically and the sampling pulses will be permanently locked to the deflection pulses. It was mentioned in Section 2.2 that the sampling pulses must be very narrow so that the desired horizontal resolution may be achieved. The narrow pulses were generated by using differentiators. For the low resolution of 120 points/line, this was quite adequate. For higher resolutions however where sampling pulses of 50ns or less are required alternative methods may be necessary. One simple way of obtaining a very narrow pulse is shown in Fig. 5.2. The duration of the pulse is given by the delay of the logic gate. For a TTL gate for instance this delay is under 10 ns. The prototype system was designed to have 2 sampling speeds, 69 ysec/sample and 138 ysec/sample. It is quite easy to have several sampling speeds by simply having several count decoders instead of two. As can be CflMERFf f : £— PULSES f Fig. 5.1 Phase locked loop scheme T T —^ T J OUT Fig. 5.2 Generation of a narrov/- pulse. seen from the photographs in Chapter 4 however a high sampling rate tends to superimpose an unacceptable sampling pattern on the picture. It would be better to find the fastest sampling rate at which the sampling pattern : disappears and use this in a final design. In the last few months, I.C and discrete component amplifiers of required characteristics to build a sample and hold to the following specifications have become available: sampling window 50 ns, hold time up to 1 usee. It was mentioned in the introduction that the sample and hold in the interface would be required to hold a sample until the next sample is taken. This is a rigid specification which is not really necessary. The computer A/D converter needs the signal -'to be available for less than 1 usee. Therefore a sample and hold with the specifications mentioned above will be quite adequate. With these modifications it should be possible to build a completely reliable system. Some of the uses to which such a system can be put will be described in the next section. 5.3 Possible Uses The field of TV bandwidth compression was one of the first areas of picture processing where computers were applied. A picture would be stored in the computer, the visual data would then be processed according to some algorithms so as to simulate various schemes proposed for reducing TV bandwidth. The processed data would be read out and displayed to see what effect the simulation has had on the quality of the picture. The interface system described in this thesis can be used as the link between the picture information and the computer. Some of the processing for the simulation of TV bandwidth compression however require adjacent samples in the computer to correspond to consecutive picture points. This is-therefore an area where arrangement of data after storage would be required. While a lot of work is still being done on TV bandwidth compression, much attention is now being paid to the relatively newer area of image processing by computer, pattern recognition. This title covers a wide area including such diverse topics as fingerprint identification, spacecraft image evaluation and biomedical image processing. The one thing these have in common is that the computer is usually only required to recognize certain characteristics of the picture information stored in it. Display of the processed data is very seldom required. Thus in the case of fingerprint identification for example, several fingerprint picture information are entered into the computer, the computer is required to compare them and find which two are identical. Weather forecasting by computer is another field of pattern recognition. The processing required to be done by the computer is similar to the one just described. Satellite weather photographs are fed into the computer, information is then extracted from this to predict the weather. Since in these and many other areas of pattern recognition there is no necessity for displaying the processed information, the interface system should be more readily acceptable as a link between the photographic information and the computer. This is because the possible loss of syn chronization during the display that was mentioned in Section 4.1 will not arise. .TITLE STORE MO. 1 (STORED .GLOBL BACK /THIS PROGRAM STORES A TV PICTURE ON TAPE THROUGH MEMORY. /EACH SAMPLE IS CONVERTED TO 11 BITS 7 0F WHICH ARE MOT /USED. 9 SAMPLES ARE PACKED IN 2 WORDS. AFTER STORING, THE /COMPUTER GOES TO ANOTHER PROGRAM TO MAKE USE OF THE DATA. / / SKPFRFr705301 CLRFRF=705302 SKPSAF=705401 CLRSAFr705402 SKPADF:705601 CLRADF=705602 CLRBUF:701104 LRS 1 1 = 640511 ADRB=701117 ADSC=701304 ALS3=640703 ALS7=643707 LRSD640501 ' • ' ' LRS7=640507 LRS5=640505 DTLA=707545 DTDF:707601 DTXA=707544 DAL 1 = 703 107 GO JMP START /THIS PART OF THE PROGRAM SETS OVER LAC (CBLK /UP THE TAPE TO ITS BEGINNING. DAC* (31 DZM* (30 . LAC (221000 DTLA WAIT DTDF JMP .-1 SEARCH LAC RBLK CMA TAD (1 TAD CBLK SMA JMP REV DAC* (30 LAC (010000 DTXA JMP WAIT REV SZA JMP INVT LAC (020000 DTXA JMP START INVT CMA TAD (1 DAC* (30 LAC (050000 DTXA JMP WAIT /THE STORING OF THE PICTURE IN /MEMORY BEGINS HERE. THE INPUT /SIGNAL IS SAMPLED AND A/D CONVERTED. /NEGATIVE SAMPLES ARE IGNORED SINCE. /THEY ONLY REPRESENT THE SYNC PART /OF THE ORIGINAL VIDEO SIGNAL. / / / START CLRBUF CLRADF LAC (163S DAC* (13 LAC (-15530 . DAC COUNT* ' CLRFRF CLRSAF SKPFRF JMP .-1 • SKPSAF JMP .-I ADSC SKPADF JMP .-1 CLRFRF CLRSAF DZM TEMP# CLA JMS SAMPLE JMP SI 0 ADD TEMP DAC TEMP ADRB SKPSAF JMP .-1 ADSC DAC HOLD# AND (004000 SAD (004000 JMP IGNORE LAC HOLD CMA TAD (1 DAL I LAC HOLD CLRSAF JMP * SAMPLE IGNORE DZM HOLD JMP LOOP / / / / / / SAMPLE LOOP / 'TOP OF PICTURE ° PULSE PRESENT? ./NO. TEST AGAIN /YES. WAIT FOR PRESENCE OF SAMPLING /PULSE THEN START SAMPLING AND /A/D CONVERTING. . -/SAMPLING PULSE PRESENT? /NO. TEST AGAIN /YES. SAMPLE AND A/D CONVERT / /THIS PART OF THE PROGRAM WILL MASK /OUT THE LAST 7 BITS OF EACH CONVERTED /SAMPLE AND PACK 9 SAMPLES IN 2 WORDS. / / 51 ALS7 AND (740300 JMS SAMPLE 52 ALS3 AMD (036000 JMS SAMPLE 53 LRS1 AND (001700 JMS SAMPLE 54 LRS5 AND (000074 JMS SAMPLE .-. 55 LRS11 , AND (000003 ADD TEMP DAC* 10 LAC HOLD LRS7 AMD (000003 . DZM TEMP JMS SAMPLE 56 ALS7 AND (740000 JMS SAMPLE 57 ALS3 AND (036300 JMS SAMPLE 58 LRS1 AND (001700 JMS SAMPLE 59 LRS5 AND (000074 ADD TEMP DAC* 10 1SZ COUNT * JMP .+2 JMS* BACK / / / / / / / / / CLA DZM TEMP ' JMS SAMPLE JMP SI /SUFFICIENT SAMPLES TAKEN? /NO. GO AND TAKE MORE /YES. GO TO SUBROUTINE TO /TRANSFER DATA TO TAPE /SUBROUTINE TO DEPOSIT DATA ON TAPE. /THE DAT IS STORED ON TAPE SO THAT /IT CAN BE PROCESSED LATER. / / WRITE LAC (-2 DAC* (30 • - • . LAC (CBLK DAC* (31 ; LAC (231300 DTLA DTDF JMP .-1 . LAC CBLK • • . • DAC SAVE# . LAC (1636 DAC* (31 LAC (-15530 DAC* (30 LAC (005000 DTXA DTDF JMP .-1 LAC (020000 DTXA JMS* BACK /EXIT TO ANOTHER PROGRAM. RBLK I CBLK 0 .END GO .TITLE DISPLAY .NO. 1 (DSPLY1) .GLOBL BACK /THIS PROGRAM PROVIDES A DISPLAY OF A TV PICTURE , /STORED IN MEMORY BY THE PREVIOUS PROGRAM, ON AN /OSCILLOSCOPE. THE SAMPLES REPRESENTING THE PICTURE /ARE UNPACKED, D/A CONVERTED AND PRESENTED TO THE /INTERFACE SYSTEM. / • / / SKPFRF=705381 CLRFRF:705332 SKPSAF=705431 • CLRSAF:705432 ALS11:640711 ALS 1:640701 -ALS5:640705 ALS7:640707 LRS3:640503 LRS7:640507 DAL 1:703107 DTLA:707545 DTDF:737601 . DTXA:707544 /SET UP FOR THE DISPLAY TAKES PLACE HERE. THE COMPUTER /WAITS FOR A 'TOP OF PICTURE' PULSE BEFORE STARTING TO /OUTPUT THE DATA. / / BACK 0 AWAY LAC (-15530 DAC COUNT* LAC (1636 DAC* (10 PLACE LAC* 10 DAC TEMP* LRS7 AND (003600 CM A TAD (1 LAC (JMP Dl DAC PLACE CLRFRF CLRSAF SKPFRF JMP .-2 SKPSAF JMP .-1 DAL 1 LAC (-3 TAD (1 SZA JMP .-2 CLRSAF LAC TEMP JMP D2 /THE UNPACKING AND DISPLAY /IN THIS SUBROUTINE. / / PROCESS TAKE PLACE DSPLY 0 AND (003600 CMA TAD (1 SKPSAF JMP .-1 DAL 1 LAC (-2 TAD (1 SZA JMP .-2 CLRSAF LAC TEMP JMP* DSPLY Dl LAC* 10 DAC TEMP LRS7 JMS DSPLY D2 LRS3 JMS DSPLY D3 ALS1 JMS DSPLY DA ALS5 JMS DSPLY D5 ALS 1 1 AND (003000 DAC KEEP# LAC* 10 DAC TEMP ALS 7 AND (000600 TAD KEEP JMS DSPLY D6 LRS7 JMS DSPLY D.7. LRS3 . JMS DSPLY D8 ALS1 JMS DSPLY D9 ALS5 JMS DSPLY ISZ COUNT JMP Dl, JMP AWAY /DISPLAY PULSE PRESENT? /NO. TEST AGAIN /YES. DISPLAY THE CORRESPONDING /SAMPLE. /ALL SAMPLES DISPLAYED? /NO. GO AND DISPLAY REMAINDER. /YES. GO BACK TO THE TOP AND /START DISPLAYING THE PICTURE /AGAIN. .END BACK BIBLIOGRAPHY 1. W.F. Schreiber and D.N. Graham, "A Digital Scanner for Computer Image Processing". Indian Institute of Technology, Dept. of E.E., April 1966. 2. Azriel Rosenfeld, "Picture Processing by Computer". New York Academic Press, 1969. 3. L.D. Harmon and K.C. Knowlton, "Picture Processing by Computer". Science Vol. 164, pp. 19-29, April 1969. 4. Tektronix, Inc., "Sampling Notes". Beaverton, Oregon, 1969. 5. Louis Nashelsky, "Digital Computer Theory". New York, John Wiley & Sons Inc., 1966. 6. O.J. Tratiak, The "SCAD", M.I.T. Research Lab of Electronics. Quarterly Progress Report 83, Oct. 15, 1966. 7. T.S. Huang and O.J. Tretiak, "Research in Picture Processing" in optical and Electro-Optical Information Processing. Chapter 3, M.I.T. Press, 1965. 8. H.U. Malmstadt and C.G. Enke, "Digital Electronics for Scientists", New York, W.A. Benjamin Inc., 1969. 9. Special Issue on "Redundancy Reduction", Proceedings of the IEEE Vol. 55, March 1967. 10. J.G. Cossalter, "A Computer Visual-Input System for the Automatic Recognition of Blood Cells", M.A.Sc. Thesis, U.B.C, August 1970. 11. E. Renschler and B. Welling, "An Integrated Circuit Phase-Locked Loop Digital Frequency Synthesizer". Application Note AN-463. Motorola Semiconductor Products. Inc. 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items