mylaboratory
PL Post Processor is a general purpose utillity that can be used on any image files. It was written becaues I needed a programme to combine the rendered images from many computers.
Monte Carlo pathtracing shows the error in it's approximation as variance in the image which results in high frequency noise. Fortunately this can be decreased by simply sampling more paths - i.e. taking more paths per pixel. Obviously this is quite expensive as four times as many samples are needed to reduce the noise by half. But, as long as the programme's random number generator is seeded correctly the image from independent runs gives different noise so it makes this method ideal for splitting between many processors. Then a final reduced-noise image can be made simply by averaging many input images.
PL Post Processor is designed to do this. It can take input images of various (mixed) formats and as long as they have the same size it will average them and display and save the result. Some extra post-processing abilities were added so the image can be modified to some extent withough having to re-run the raytracer. The programme also supports the High Dynamic Range (HDR) format that I created so that exposure effects can be applied without the limits of standard 256 values per colour channel.
The latest version is 1.0.0 - 1st March 2006.
The HDR > SDR Output window displays the resulting image after processing.
The Files tab controls loading and saving of images.
(a) Tabs for file loading or processing.
(b) Directory selection.
(c) File selection. All highlighted files are loaded and averaged together.
(d) High dynamic range and standard range output files and browse boxes.
(e) Buttons to load image files and save the HDR and SDR files.
(f) Load progress indicator.
The Post Processing tab applies various functions to the data and controls the order in which they are performed.
(a) The 'Effect order' group allows the order of the functions to be changed and switched on or off. The buttons at the top move the currently selected item up or down and double clicking on a line toggles it's state.
(b) The exposure of the image can be changed in either tab stops (*2^n) or by a plain multiplication factor. This works best on HDR images because SDR images will loose some of their range.
(c) Changes the gamma factor of the image. Each channel is taken to the power of the gamma factor. It is useful to bringing out detail at either end of the brightness range.
(d) Multiplies the three colour channels by the three factors to changet he colour balance.
(e) Multiplies the hue, saturation and value colour descriptions by factors and can also rotate the hue.
(f) Applies a cellshading-like effect to the image. The colour channels are quantised to n+1 values (the clamp feature does not do anything at the moment but is designed to take account of the HDR nature of images).
(g) Gaussian blurs the image with a defined full width at half maximum and nxn filter window.
(h) Makes bright regions of the image spread out slightly to form a glow-like effect. The threshold is the 'value' size that triggers the effect on a pixel. The FWHM and size have the same meanings as for glow and the cheat factor adds an extra multiplication value to enhance the effect.
(i) Attempts to detect the edges of objects in the image. The detection method set the component to check for changes (eg hue, red channel etc), the colour method selects how the edge colour will be changed (eg replace, invert etc). The difference sets the minimum difference between pixels that triggers an edge and the width sets the width of the line. Finally colour sets the colour to draw edges (if replace is on) and blueprint replaces everything that isn't an edge by a colour.
The effects shown are (left to right, top to bottom):
(1) Original image
(2) Gamma correction of 0.5
(3) Saturation to 0
(4) RGB levels, (3, 0.5, 1)
(5) 120 degree hue rotation
(6) Glow effect
(7) Exposure up 3/4 of a stop followed by cellshading of the colours to three values per channel
(8) Edge detection using a difference of 0.05 of 'value' (HSV colour space) then replacing the colour with a grey corresponding to average of the pixel followed by replacing the rest with black (for a blackboard effect)
(9) +3/4 aperture stops, 1.5 gamma correction, gaussian blur to remove noise, 270 degree hue rotation, 3 colour per channel cellshade, RGB edge detection (threshold 0.3) replaced with black
