Get Adobe Flash player
T h e h s a n d l o w s h ig by Robin Palmer H istogram displays for video analysis probably followed those found in common computer graphics programs such as Adobe Photoshop. This can be very useful in finding video signal problems that would be difficult or impossible to see with a conventional waveform display. The histogram is a way of showing statistical results as a bar-chart. The range of possible signal values goes on the x-axis whilst a count of each is presented as the height of each bar. The most frequently occurring values show as peaks. When this technique is applied to live video, all the count heights are automatically scaled to whichever is the highest of the available space. This is done to make the whole display a sensibly scaled size. This normalisation is why a live histogram can appear to dance up and down as the scaling is varied to match the changes in the peak values. Generating a live histogram involves quite a lot of computation. There is a bank of counters, one for every possible pixel value. As all pixels in each frame are sampled, the value- specific counter is incremented. At the end of each TV frame all the counters are scanned to find the highest and o f h i s t o g r a m s the scaling factor for 100% computed. Each counter is then read out and multiplied by the scaling factor to build the height of each bar. This normalised chart is then drawn by the systems graphics in real time, frame by frame. The histogram display can be applied to luma (brightness) or to the red, green and blue signal components. These are able to show some common potential faults. A high percentage at either end of the graphs can show blacker-than-black or whiter-than-white arising from gamut errors. In 8 bit video these values are 0-15 for ultra-black and 236–255 for over-white. A small percentage of these excursions caused by overshoots are normal and quite allowable. If the graph resembles a comb or teeth with fixed spacing gaps between dancing live values then these are quantisation errors. Missing code values will show as contours on areas of gentle graduation like a piece of sky in shot. This can be very distracting particularly with fades. If the ‘teeth’ are regular, they are most likely to have come for transcoding from a lower bit range, contrast resolution or processing. Single fixed gaps can come from hardware faults like troublesome capture cards with missing codes. If the histogram is used in the edit suite, the overall contrast range can be monitored. If the contrast is expanded because the high-lights are too dim or the blacks look grey, the range can be optimised. The contrast expansion could introduce quantisation gaps although some editing packages can interpolate to avoid this problem. The best policy is to shoot with a full exposure range in the first place. This can be ensured by observing histograms as well as waveform monitoring whilst on-set. RGB histograms can be used similarly to luma but are not easily applied to colour-matching tasks. The exception is in 3D where both left and right eye histograms in RGB can show the discrepancies between the camera adjustments that need attention. To help comparisons, a line graph of just the peak values can be shown instead of a bar chart. This then allows a readable superimposition of left and right eye data. Nearly all modern video instruments feature histogram capabilities. They can also be a built-in facility on normal video and viewfinder screens. For instance, the Cel-Soft Reel- Check Solo uses PC software image processing that can show multiple histograms and many other displays including SD/HD/2K and even 4K. Similarly for 3D production or post, the Cel-Scope 3D software suite can superimpose left and right eye luma or RGB histograms for real-time comparison and matching as line graphs or bar-charts. Robin Palmer is Managing Director of Cel-Soft and is habitually involved with solutions for 3D & TV technology. 36 | TV-BAY MAGAZINE TV-BAY071NOV12.indd 36 06/11/2012 18:09