Abstract: A bad pixel correction (BPC) algorithm that can be implemented on the image sensor chip is provided for detecting and correcting defective pixels in a digital color image sensor. Gradients of neighboring pixels in at least one other color plane than the color plane of a current pixel and a range of sensor values from neighboring pixels in the same color plane as the current pixel are determined. If the sensor value of the current pixel is outside of the range by a threshold amount that is calculated using one or more of the gradients, the current pixel is determined to be a defective pixel, and replaced using the sensor values of the neighboring pixels in the same color plane.

Abstract: A digital image system and method for combining bad pixel correction and demosaicing in a single process is provided by interpolating sensor values for pixels immediately spatially adjacent to the current pixel being examined to detect defective pixels, and using the interpolated values for demosaicing. If the sensor value of the current pixel being examined is outside of a range of sensor values determined from the interpolated values by more than a configurable threshold amount, the current pixel is considered defective, and replaced using an estimated value from the neighboring pixels.

Abstract: Previously-unused slots in a Huffman code table associated with a Joint Photographic Experts Group (JPEG) image file are associated with various quantization matrices (Q matrices) that are used to quantize data blocks of the JPEG image file. Huffman codes associated with the various Q matrices permit the particular Q matrix used to quantize a given data block to be signaled by a decoder as an end-of-block (EOB) code. The EOB codes and the Huffman code table are sent with the JPEG image file. Upon decoding of the image file, a standard JPEG decoder reads each of the EOB codes as a standard JPEG EOB code and does not vary the Q matrix. A modified decoder reads from each of the EOB codes which Q matrix was used to encode each particular data block of the image and uses that Q matrix to dequantize the data block.

Abstract: An adaptive demosaicing method is disclosed that interpolates images based on color edge detection and neighborhood voting. The adaptive demosaicing algorithm uses a voting scheme to determine the direction of interpolation at each missing luminance pixel location. Each color plane votes either horizontal or vertical based on a comparison between the vertical and horizontal components of the degree of change (i.e., gradient, Laplacian or other measure of the degree of change) in that color plane. Votes are counted from the neighborhood pixels as well as from measurements taken at the pixel location itself. Once the luminance plane is fully interpolated, the chrominance planes are filled in by a simple bilinear or median interpolation of difference chrominance values. Enhancements to the adaptive demosaicing algorithm permit adaptive smoothing and sharpening.

Abstract: A system and method for processing mosaiced images utilizes a compression-aware demosaicing process that takes into consideration a subsequent compression process. The compression-aware demosaicing process is performed using a compression-considered demosaicing operator that incorporates a color space conversion operator and a frequency-based transformation operator, which are typically associated with the compression process. Consequently, the overall efficiency of the system and method is significantly increased. Furthermore, the compression-aware demosaicing process produces artifacts that complement the artifacts produced by the subsequent compression process such that the artifacts are less visible in the final color images, which increases the quality of the final color images.

Abstract: A sensor includes an array of photodetectors each generating an output signal of pixel data indicative of incident light intensity. This pixel data is read out from the array one line at a time and stored in a line buffer. A bad pixel processor includes a first buffer that stores pixel data obtained from the line buffer for a certain pixel in a currently read out line and pixel signal light data for pixels adjacent to the certain pixel. An included second buffer stores features that are indicative of whether the pixels in a previously read out line were identified as bad pixels. Using the information in the first and second buffers, the processor identifies whether the certain pixel is a bad pixel. In one operation, the processor precludes any finding of the certain pixel as being a bad pixel if the features in the second buffer indicate that a pixel in the previous line that is adjacent to the certain pixel in the current line was identified as a bad pixel.

Abstract: A system and method is provided for processing a demosaiced image using a color aliasing artifact reduction (CAAR) algorithm in order to reduce color aliasing artifacts. The CAAR algorithm computes the L level wavelet transform for the demosaiced color planes R, G and B. Thereafter, the CAAR algorithm estimates the correct color value at each pixel location for the colors not associated with that pixel location. For example, to determine the green value at red pixel locations, the CAAR algorithm performs an inverse wavelet transform using the green approximation signal and the red detail signals. This process is repeated for each of the colors (e.g., green values at blue pixel locations, red values at green pixel locations, etc.). In addition, the CAAR algorithm performs an inverse wavelet transform on each of the color planes themselves, so that the pixel values of the color associated with each pixel location are not altered.