Abstract: A convolution operation method is provided for performing a convolution operation on an input feature map to generate a corresponding output feature map, wherein the input feature map is divided into a plurality of input data blocks, and the convolution operation method includes: dividing each of the input data blocks into a plurality of non-overlapping areas, wherein there is an overlapping area between any two adjacent input data blocks; storing the non-overlapping areas of each input data block into a respective non-overlapping storage space in a cache; generating each input data block according to the area corresponding to each input data block stored in the non-overlapping storage spaces; and performing a convolution operation on the plurality of generated input data blocks to generate the output feature map.

Abstract: An image anti-aliasing method comprises following steps: receiving an aliasing image; wherein the aliasing image includes a plurality of source pixels; generating a plurality of hash values corresponding to the plurality of source pixels respectively; and performing a filtering processing or a filter generating procedure on the aliasing image according to the plurality of hash values. Each of the plurality of hash values is generated by the following steps: selecting one source pixel from the aliasing image and selecting a window containing the selected source pixel; determining an aliasing strength according to the grayscale values of all the source pixels of a row and a column at which the selected source pixel is located in the window; and determining a hash value of the selected source pixel according to the aliasing strength.

Abstract: An image anti-aliasing method comprises following steps: receiving an aliasing image; wherein the aliasing image includes a plurality of source pixels; generating a plurality of hash values corresponding to the plurality of source pixels respectively; and performing a filtering processing or a filter generating procedure on the aliasing image according to the plurality of hash values. Each of the plurality of hash values is generated by the following steps: selecting one source pixel from the aliasing image and selecting a window containing the selected source pixel; determining an aliasing strength according to the grayscale values of all the source pixels of a row and a column at which the selected source pixel is located in the window; and determining a hash value of the selected source pixel according to the aliasing strength.

Abstract: A technique for graphics processing which processes an image in units of blocks is disclosed. A graphics processing device has a buffer, an indicator register and an arithmetic logic unit (ALU). The buffer is provided to buffer a plurality of source blocks. The indicator register includes at least one instruction pointer. The ALU uses the instruction pointer to read source pixels in a first source block of the plurality of source blocks and adjacent source pixels in at least an adjacent source block with respect to the first source block from the buffer, to generate a first target block.

Abstract: A technique for graphics processing, which completes graphics processing of an image loaded from a system memory by performing a series of slice processing steps. A device for graphics processing has an internal vector dynamic memory for buffering slices of pixel data loaded from the system memory. The internal vector dynamic memory has a first buffer for buffering non-overlapped pixel data, which is not reused in a next slice processing step and a second buffer for buffering overlapped pixel data, which is reused in the next slice processing step.

Abstract: A technique for graphics processing which processes an image in units of blocks is disclosed. A graphics processing device has a buffer, an indicator register and an arithmetic logic unit (ALU). The buffer is provided to buffer a plurality of source blocks. The indicator register includes at least one instruction pointer. The ALU uses the instruction pointer to read source pixels in a first source block of the plurality of source blocks and adjacent source pixels in at least an adjacent source block with respect to the first source block from the buffer, to generate a first target block.

Abstract: A technique for graphics processing, which completes graphics processing of an image loaded from a system memory by performing a series of slice processing steps. A device for graphics processing has an internal vector dynamic memory for buffering slices of pixel data loaded from the system memory. The internal vector dynamic memory has a first buffer for buffering non-overlapped pixel data, which is not reused in a next slice processing step and a second buffer for buffering overlapped pixel data, which is reused in the next slice processing step.

Abstract: A method for generating and employing a camera noise model, performed by a processing unit, is introduced to at least contain the following steps. A first frame is obtained by controlling a camera module via a camera module controller. A camera noise model is generated on-line according to the content of the first frame, which describes relationships between pixel values and standard deviations. A second frame is obtained by controlling the camera module via the camera module controller. The content of the second frame is adjusted using the camera noise model, and the second frame, which has been adjusted, is stored in a frame buffer.

Abstract: A method for correcting bad pixels, performed by a processing unit, at least contains: reading a block of a frame; labeling each pixel of the block as a good pixel, a weak pixel, or a bad pixel; detecting a weak-pixel pair from the block, which includes two pixels labeled as weak pixels; re-labeling the pixels of the weak-pixel pair as bad pixels converted from weak pixels; and correcting the pixels labeled as the bad pixels, and the bad pixels converted from weak pixels.

Abstract: A method for correcting bad pixels, performed by a processing unit, at least contains: reading a block of a frame; labeling each pixel of the block as a good pixel, a weak pixel, or a bad pixel; detecting a weak-pixel pair from the block, which includes two pixels labeled as weak pixels; re-labeling the pixels of the weak-pixel pair as bad pixels converted from weak pixels; and correcting the pixels labeled as the bad pixels, and the bad pixels converted from weak pixels.

Abstract: A method for generating and employing a camera noise model, performed by a processing unit, is introduced to at least contain the following steps. A camera noise model is provided. A first frame is obtained by controlling a camera module via a camera module controller. A blending ratio corresponding to each pixel value of the first frame is generated according to the camera noise model, the pixel value of the first frame and a corresponding pixel value of a second frame. A third frame is generated by fusing each pixel value of the first frame with the corresponding pixel value of the second frame according to the blending ratio. A de-nosing strength for each pixel value of the third frame is adjusted according to the blending ratio. Each pixel value of the third frame is adjusted using the corresponding de-nosing strength.

Abstract: A method for generating and employing a camera noise model, performed by a processing unit, is introduced to at least contain the following steps. A camera noise model is provided. A first frame is obtained by controlling a camera module via a camera module controller. A blending ratio corresponding to each pixel value of the first frame is generated according to the camera noise model, the pixel value of the first frame and a corresponding pixel value of a second frame. A third frame is generated by fusing each pixel value of the first frame with the corresponding pixel value of the second frame according to the blending ratio. A de-nosing strength for each pixel value of the third frame is adjusted according to the blending ratio. Each pixel value of the third frame is adjusted using the corresponding de-nosing strength.

Abstract: A method for generating and employing a camera noise model, performed by a processing unit, is introduced to at least contain the following steps. A first frame is obtained by controlling a camera module via a camera module controller. A camera noise model is generated on-line according to the content of the first frame, which describes relationships between pixel values and standard deviations. A second frame is obtained by controlling the camera module via the camera module controller. The content of the second frame is adjusted using the camera noise model, and the second frame, which has been adjusted, is stored in a frame buffer.