Automated monitoring of digital video image quality

A system for automated monitoring of rectangular visual artifacts (MBAs) in block-compressed digital videos presented as analog video, comprising:

• a digitizer digitizing analog video into at least one image memory having an image size;
• an FMG builder
• estimating video and compression block formats based on said image size, said at least one image memory, and video parameters;
• computing exclusion areas, of a digital image in said at least one image memory, that are excluded from further use based on said digital image size, said at least one image memory, and said video parameters;
• and
• determining horizontal edges and vertical edges of potential MBAs;
• a junction detector
• locating a set of junctions where said horizontal edges and said vertical edges cross;
• deleting, from said set of junctions, close junctions;
• computing a score for each junction in said set of junctions based on neighboring junctions;
• and
• storing locations of said junctions and said scores in a temporal sequence of said potential MBAs of feature memories;
• a projection detector
• creating an array of horizontal projections of said horizontal edges;
• suppressing peaks in said horizontal projections based on positions of said horizontal projections with respect to said horizontal edges, and based on amplitudes of said horizontal projections;
• and
• storing said array of horizontal projections and the unsuppressed peaks, in a temporal sequence of said potential MBAs of said feature memories;
• a temporal filter applied to said temporal sequence to accept or reject said temporal sequence;
• and
• a decision network detecting the MBAs in said temporal sequence and providing the MBAs by a computer signal.

A method for automated monitoring of rectangular visual artifacts (MBAs) in block-compressed digital videos presented as an analog signal, comprising the steps of:

• (a) accessing information from a digital video distribution to monitor a video channel;
• (b) selecting, decoding, and decompressing the information into the analog signal;
• (c) digitizing the analog signal
• (d) storing a digitized image of the analog signal into at least one image memory;
• (e) obtaining a digitized image size from the digitized image;
• (f) accessing video parameters;
• (g) estimating dimensions and locations of possible MBAs from the digitized image size, the video parameters, and the black lines in the at least one image memory;
• (h) computing exclusion areas of the digital image to exclude from further use, based on the digitized image size, the video parameters, and black lines on the edges of the at least one image memory;
• (i) creating a flexible macroblock grid (FMG) to indicate possible positions of potential MBAs based on the estimated dimensions, the estimated locations, and the exclusion areas;
• (j) for each of the digitized images, determining locations of vertical edges of the potential MBAs based on the FMG;
• (k) differentiating digital image intensity/color of the digitized images to detect horizontal edges and vertical edges and suppress other image structure to create an edge image;
• (l) transforming the edge image into an edge map to remove low contrast edges from consideration;
• (m) locating junctions where the horizontal edges and vertical edges cross in the edge map, the junctions being in proximity to the locations and the dimensions in the FMG;
• (n) removing or suppressing the close junctions;
• (o) assigning a score to each of the unsuppressed junctions based neighboring junctions and the dimensions;
• (p) storing the locations and the scores of the remaining junctions in feature memories as a function of time;
• (q) vertically differentiating the digitized image to create a horizontal edge;
• (r) integrating the horizontal edge to create a horizontal projection array;
• (s) updating the horizontal projection array to remove peaks that are further than a preselected distance from the locations based on the FMG;
• (t) modifying the updated horizontal projection array to remove the peaks that are less than a preselected value;
• (u) searching the modified horizontal projection array to find the peaks that are near horizontal edges of the locations from the FMG;
• (v) storing the horizontal projection array and the near peaks in the feature memories as a function of time;
• (w) temporally filtering a sequence of the junctions and the projections in the feature memories to discard sequences of the potential MBAs that are shorter than a preselected minimum time or longer than a preselected maximum time;
• (x) reporting sequences that are not filtered out by said step of temporally filtering to a decision network;
• (y) determining, by the decision network, whether there are MBAs in the digitized image based on the junctions and the projections data in the filtered feature memories and decision parameters;
• and
• (z) raising signals based on occurrence of the MBAs.