Video filtering for stereo images

An image processing system for generating at least one output image associated with an output viewpoint from an input image associated with an input viewpoint through a depth-dependent transformation;

• the images being represented as an input pixel array and output pixel array, respectively;
• the image processing system including:
• an input for receiving the input image, where the input image includes for each input pixel an associated input pixel value and input pixel depth, wherein a pixel being indicated by an x-coordinate and a y-coordinate;
• each input pixel being associated with a respective reconstruction filter footprint;
• a video processor operative to create output pixels of the output image by:
• for each input pixel, transforming the input pixel to a transformed input pixel as a function of the associated input pixel depth and transforming the associated filter footprint to an transformed filter footprint as a function of the associated input pixel depth;
• and
• performing a reconstruction filtering operation on a plurality of the transformed input pixels using the transformed filter footprints;
• and
• an output for providing the output image for subsequent rendering, wherein the video processor is operative to determine which transformed filter footprints are fully not-occluded and the video processor is operative to perform the reconstruction filtering operation based on fully not-occluded transformed filter footprints, wherein the video processor being operative to sequentially process input pixels of a row in a direction opposite from a displacement from the input viewpoint to the output viewpoint along the x-axis;
• the processing including:
• maintaining an x-coordinate extent that indicates for already processed input pixels with respect to a predetermined start position a furthest x-coordinate already occluded by at least one transformed input pixel, where the furthest x-coordinate is a highest x-coordinate if the processing direction is from left-to-right and a lowest x-coordinate if the processing direction is right-to-left;
• and determining that a transformed input pixel is occluded from the output viewpoint if processing of the associate input pixel does not result in increasing the x-coordinate extent.