A second generation image coding technique using human visual system based segmentation

Image Segmentation Using Human Visual System Properties With Applications In Image Compression

“…In image segmentation, the pixels in an image are divided into mutually exclusive spatial regions based on some criteria. The criteria used could be as simple as the similarity of the pixel gray levels (yielding flat image segments) [3,6], or the criteria could be more complex, such as how well the pix els fit a given planar model (facet-based segmentation) [82], a two-dimensional polyno mial model [87], or a statistical model (texture-based segmentation). After segmenta tion, the image consists of regions separated by contours.…”

“…In most segmentation-based compression schemes, the shapes of the image segments are represented by encoding the segment boundaries. These boundaries may be coded by approximating them with straight lines and circle segments and then coding the information describing this approximation [82], or by a more simple approach, such as coding a binary image describing where segment boundaries are located in the image [3,6]. The interiors of the segments are represented by encoding, for example, the coefficients in the polynomial models describing each segment, or for fiat segments, the average gray level of the pixels in each segment.…”

“…The initial segmentation algorithm uses a variation of centroid-linkage region growing [5], and is based on a technique presented in [6,95]. With centroid-linkage region growing, the image pixels are scanned in a raster fashion.…”

“…In this section we discuss a filtering technique for the elimination of visually insignificant segments from a segmented discrete gray level image [6]. This filtering operation, which we refer to as post-segmentation filtering, takes advantage of HVS properties relative to contrast sensitivity.…”

“…A third difference in our compression technique involves the representation we use for the image segment shapes. Other segmentation-based gray level image compression techniques have typically represented the image segments by encoding the segment boundaries [2,6,7]. We propose the use of skeletons generated using mathematical morphology to represent the segment shapes.…”