Abstract: Applying a mechanism of image signal processing and color-space-conversion to convert the captured Green components to be Y, luminance components for only those pixels having raw Green data without interpolation, and to convert the Blue components to be U, chrominance components for only those pixels having raw Blue data, and Red components to be V, chrominance components for only those pixels having raw Red data. These converted YUV components are input to a predetermined video compression codec for reducing the intra- and inter-frame redundant information.
Type:
Grant
Filed:
December 11, 2007
Date of Patent:
April 23, 2013
Assignee:
Taiwan Imagingtek Corporation
Inventors:
Chih-Ta Star Sung, Eckehard Steinbach, Yin-Chun Blue Lan, Hu Chen, Mingzhe Sun
Abstract: A method of compressing an image by saving the compressed color components into multiple temporary buffers with each time slot saving the color component to the different buffer. A method deciding when to shift the compressed color components to another device is applied to avoid underflow and overflow of the output buffer.
Abstract: The differential value of the adjacent pixels is calculated firstly and is coded by a VLC coding. The VLC coding includes codes representing the Quotient and Remainder with a marker bit inserted in between. The divider is predicted with no code in the coded data stream. If three pixel components are presented in the same clock time, three VLC encoders and three VLD decoders are applied to encode and decode one pixel at a time. During encoding, both Remainder and Quotient of the same pixel component are encoded in parallel followed. During decoding, both Remainder and Quotient of the same pixel component are decoded in parallel and the results of the first pixel component are used a reference to decode the second pixel component which adopts the same decoding procedure and the decoded results of the second pixel component is used as reference to decode the third pixel component.
Type:
Grant
Filed:
November 3, 2006
Date of Patent:
May 31, 2011
Assignee:
Taiwan Imagingtek Corporation
Inventors:
Chih-Ta Star Sung, Yin-Chun Blue Lan, Shi-Ho Tien
Abstract: The present invention provides a new semiconductor Read-Only Memory, ROM, which stores more than one bit per cell. The potential of multiple threshold voltages combined with the potential multiple ratios of device channel width and length makes an ROM cell store multiple bits feasible. An N-type or a P-type MOS device of the standard CMOS process or a flat-cell mask ROM process are operable devices and processes in the design of this multi layer cell ROM. The ROM cell with smaller size is implemented to represent the LSB bits, while the larger size ROM cell is to represent the MSB bits.
Type:
Grant
Filed:
December 3, 2004
Date of Patent:
September 15, 2009
Assignee:
Taiwan Imagingtek Corporation
Inventors:
Chih-Ta Star Sung, Thomas Chang, Ing-Ruey Liaw
Abstract: An image compression method for compressing groups of pixel data, including the following steps: MSB portions and LSB portions of the pixel data are compressed separately. Further, the LSB portions are grouped according to their associated MSB portions so that similar pixels are compressed together to gain higher a compression ratio. Some LSB portions are truncated to ensure a compression ratio. When performing compression, enhanced variable-length coding is adopted that predicts more accurate dividers by considering weighted combination of adjacent pixels. The quotient of each groups of compression portion is predicted and coded accordingly.
Type:
Grant
Filed:
November 26, 2004
Date of Patent:
December 16, 2008
Assignee:
Taiwan Imagingtek Corporation
Inventors:
Chih-Ta Star Sung, Chih-Sheng Cheng, Yin-Chun Lan
Abstract: The present invention provides method of a variable length coding for data compression. The first coding algorithm is applied to encode the data with the value less than half of the calculated divider and another coding algorithm encoding the subtracted value of the data. Another algorithm applies at least two predetermined “marker bits” to indicate the range of data with concatenated binary code encoding the shifted value of the data. Another algorithm deciding the boundary separating the first range and other ranges of the data and encode the data by applying at least two different algorithms accordingly.