Image analysis – Image compression or coding – Pyramid – hierarchy – or tree structure
Reexamination Certificate
2001-04-23
2004-10-12
Do, Anh Hong (Department: 2624)
Image analysis
Image compression or coding
Pyramid, hierarchy, or tree structure
C382S232000, C382S239000
Reexamination Certificate
active
06804403
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates generally to image coding and more particularly to compression and decompression of scalable and content-based, randomly accessible digital still images.
BACKGROUND OF THE INVENTION
The fast growth of Internet and digital multimedia applications has created a consistent and growing demand for new image coding tools that reduce the usually large and cumbersome raw image data files into a compressed form. Compactness of the resulting bit-stream, however, is no longer the only requirement asked of developers when devising new coding tools. End users and their applications are increasingly demanding features like scalability, error robustness and content-based accessibility.
Photographs or motion picture film are two-dimensional representations of three-dimensional objects viewed by the human eye. These methods of recording two-dimensional versions are “continuous” or “analog” reproductions. Digital images are discontinuous approximations of these analog images made up or a series of adjacent dots or picture elements (pixels) of varying color or intensity. On a computer or television monitor, the digital image is presented by pixels projected onto a glass screen and viewed by the operator. The number of pixels dedicated to the portrayal of a particular image is called its resolution i.e. the more pixels used to portray a given object, the higher its resolution.
A monotone image—black and white images are called “grayscale”—of moderate resolution might consist of 640 pixels per horizontal line. A typical image would include 480 horizontal rows or lines with each of these containing 640 pixels per line. Therefore, a total of 307,200 pixels are displayed in a single 640×480 pixels image. If each pixel of the monotone image requires one byte of data to describe it (i.e. either black or white), a total of 307,200 bytes are required to describe just one black and white image. Modern gray scale images use different levels of intensity to portray darkness and thus use eight bits or 256 levels of gray. The resulting image files are therefor correspondingly larger.
For color images, the color of each pixel in an image is typically determined by three variables: red (R), green (G), and blue (B). By mixing these three variables in different proportions, a computer can display different colors of the spectrum. The more variety available to represent each of the three colors, the more colors can be displayed. In order to represent, for example, 256 shades of red, an 8-bit number is needed. The range of the values of such a color is thus 0-255. The total number of bits needed to represent a pixel is therefor 24 bits—8 bits each for red, green, and blue, commonly known as RGB888 format. Thus, a given RGB picture has three planes, the red, the green, and the blue, and the range of the colors for each pixel in the picture is 0-16.78 million, or R×G×B=256×256×256. A standard color image of 640×480 pixels therefor, requires approximately 7.4 megabits of data to be stored or represented in a computer system. This number is arrived at by multiplying the horizontal and vertical resolution by the number of required bits to represent the full color range—640×480×24=7,372,800 bits.
Standard, commonly available hardware, while increasingly fast and affordable, still finds files of this size slow and unwieldy. This is especially true in the case of interactive applications and Internet use. Interactive applications demand very fast multi-directional processing of multimedia data. Given their persistently large size, image files have been a rate limiting factor in the development of realistic, interactive computer applications. In the case of the Internet, end-users and applications are further limited by the slow pace of modems and other transmission media. For example, the amount of information currently capable of being transmitted over a telephone line in the interval of one second is restricted to 33,600 bits-per second due to the actual wires and switching functions used by the typical telephone company. Therefore, a single, full color RGB888 640×480 pixel page, with its 7,372,800 bits of data would take approximately three and one half minutes to transfer at this baud rate.
Many methods of compressing image data exist and are well known to those skilled in the art. Some of these methods are as “lossless” compression; that is, upon decoding and decompressing they restore the original data without any loss or elimination of data. Because their relative reduction ratios are small however, these lossless techniques cannot satisfy all the current demands for image compression technologies. Other compression methods exist that are nonreversible and known as “lossy”. These nonreversible methods can offer considerable compression, but do result in a loss of data. In image files, the high compression rates are actually achieved by eliminating certain aspects of the image, usually those to which the human eye has limited or no sensitivity. After coding, an inverse process is performed on the reduced data set to decompress and restore a reasonable facsimile of the original image. Lossy compression techniques may also be combined with lossless methods for a variable mix of data compression and image fidelity.
Compactness of a compressed bit-stream is usually measured by the size of the stream in comparison to the size of the corresponding uncompressed image data. A quantitative measure of the compactness is the compression ratio, or alternatively, the bit-rate where:
compression ratio=(total bytes of the original raw image data)/(total bytes required for compressed image)
and
bit-rate=(total bytes required for decompression)/(pixel number of the original image)
In general, the higher the compression ratio (or the lower the bit-rate), the higher the compactness of a compressed bit-stream. Compactness has been always a primary concern for all data compression techniques.
One of the most popular formats for compressed image files is the GIF format. GIF stands for “Graphic Image Format”, and was developed by Compuserve to provide a means of passing an image from one dial-up customer to another, even across different computer hardware platforms. It is a relatively old format, and was designed to handle a palette of 256 colors—8 bit as opposed to 24 bit color. When developed, this was near state of the art for most personal computers.
The “GIF” format uses an 8 bit Color Look Up Table (sometimes called a CLUT) to identify color values. If the original image is an 8 bit, gray-scale photo, then the “GIF” format produces a compressed lossless image file. A gray scale image typically has only 256 levels of gray. The operative compression is accomplished by the “Run Length Encoding” (RLE) mechanism of compressing the information while saving a GIF file. If the original file were a 24 bit color graphic image, then it would first be mapped to an 8 bit CLUT, and then compressed using RLE. The loss would be in the remapping of the original 24 bit (16.7 million) colors to the limited 8 bit (256 colors) CLUT. RLE encoding would reproduce an uncompressed image that was identical to the remapped 8 bit image, but not the same as the original 24 bit image. RLE is not an efficient way of compressing an image when there are many changes in the coloration across a line of pixels. It is very efficient when there are rows of pixels with the same color or when a very limited number of colors is used.
The other de facto standard of still image formats is the JPEG format. JPEG stands for Joint Photographic Experts Group. JPEG uses a lossy compression method to create the final file. JPEG files can be further compressed than their GIF relations, and they can maintain more color depth than the 8 bit table used in the GIF format. Most JPEG compression software provides the user with a choice between image quality, and the amount of compression. At compression ratios of 10:1 most images look very much like the or
Qu Li
Simon Brent
Wang Meng
Yang Xue Dong
Digital Accelerator Corporation
Do Anh Hong
Morrison & Foerster / LLP
LandOfFree
Region-based scalable image coding does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Region-based scalable image coding, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Region-based scalable image coding will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3267281