Abstract: An optimization method includes determining first information related to a difference between the color filter spectrum and the color spectrum or a difference between an image transformed by the color transformation matrix and a ground truth (GT) image in the reset color space, determining second information representing smoothness of the color filter spectrum, third information representing a transmittance of the color filter spectrum, calculating a cost value based on the first information, the second information, and the third information, compare the calculated cost value and a threshold, and updating one of the filter spectrum information and the color transformation matrix in response to the calculated cost value being equal to or greater than the threshold.

Abstract: A method and an electronic device with image registration are provided. The method includes generating a first optical flow between a first partial image and a second partial image, which are captured by respective first and second cameras using an optical flow estimation model; generating disparity information between the first partial image and a third partial image, captured by a third camera, based on depth information of the first partial image generated using the first optical flow; and estimating a second optical flow between the first partial image and the third partial image based on the generated disparity information for generating a registration image.

Abstract: An electronic device includes an image sensor including a color filter having a plurality of color channels, a memory storing a plurality of color conversion matrices and instructions, and a processor. The processor is configured to obtain noise information of a color image captured by the image sensor, select a target matrix from among the plurality of the color conversion matrices based on the obtained noise information, and generate a color converted image by applying the selected target matrix to the color image.

Abstract: Disclosed is a color transform method performed by an electronic device that includes generating an initial color transformed image by performing color transform on a raw image, determine noise amplification degrees indicating degrees to which noise included in the raw image is amplified by the color transform, processing the determined noise amplification degrees, and generating a final color transformed image by filtering the generated initial color transformed image using the processed noise amplification degrees and at least one of the raw image or luminance information of the raw image.

Abstract: A method of learning a parameter of a sensor filter and an apparatus for performing the method are provided. The learning method may include performing a simulation on a target image for each spectrum of a sensor filter, obtaining an output value by inputting the simulated image to a vision model for a vision task, and learning a parameter of the sensor filter based on a loss between a label of the vision model and the output value of the vision model.

Abstract: A target tracking method and apparatus is provided. The target tracking apparatus includes a memory configured to store a neural network, and a processor configured to extract feature information of each of a target included in a target region in a first input image, a background included in the target region, and a searching region in a second input image, using the neural network, obtain similarity information of the target and the searching region and similarity information of the background and the searching region based on the extracted feature information, obtain a score matrix including activated feature values based on the obtained similarity information, and estimate a position of the target in the searching region from the score matrix.

Abstract: An optimization method includes determining first information related to a difference between the color filter spectrum and the color spectrum or a difference between an image transformed by the color transformation matrix and a ground truth (GT) image in the reset color space, determining second information representing smoothness of the color filter spectrum, third information representing a transmittance of the color filter spectrum, calculating a cost value based on the first information, the second information, and the third information, compare the calculated cost value and a threshold, and updating one of the filter spectrum information and the color transformation matrix in response to the calculated cost value being equal to or greater than the threshold.

Abstract: An apparatus with image processing includes: one or more processors configured to: generate a full sampling image by demosaicing a first color filter array (CFA)-based input image; based on a trained color conversion model trained for color conversion and noise suppression of the full sampling image, determine a bias and a conversion matrix corresponding to each pixel of the full sampling image; and based on the conversion matrix and the bias, convert a color space of the full sample image, corresponding to the first CFA, into a color space corresponding to a second CFA.

Abstract: A method and apparatus for color space conversion are provided. An apparatus for converting a color space includes at least one processor configured to convert an original full image of an original color space to a temporary image of a target color space, estimate a color space mapping parameter between the original color space and the target color space, the color space mapping parameter corresponding to the original full image, obtain a residual vector of the target color space based on the temporary image and the color space mapping parameter, convert the residual vector to a residual image of the target color space, and obtain a target full image of the target color space by combining the residual image with the temporary image.

Abstract: An electronic device includes an image sensor including a color filter having a plurality of color channels, a memory storing a plurality of color conversion matrices and instructions, and a processor. The processor is configured to obtain noise information of a color image captured by the image sensor, select a target matrix from among the plurality of the color conversion matrices based on the obtained noise information, and generate a color converted image by applying the selected target matrix to the color image.

Abstract: A target tracking method and apparatus is provided. The target tracking apparatus includes a memory configured to store a neural network, and a processor configured to extract feature information of each of a target included in a target region in a first input image, a background included in the target region, and a searching region in a second input image, using the neural network, obtain similarity information of the target and the searching region and similarity information of the background and the searching region based on the extracted feature information, obtain a score matrix including activated feature values based on the obtained similarity information, and estimate a position of the target in the searching region from the score matrix.

Abstract: Disclosed is a color transform method performed by an electronic device that includes generating an initial color transformed image by performing color transform on a raw image, determine noise amplification degrees indicating degrees to which noise included in the raw image is amplified by the color transform, processing the determined noise amplification degrees, and generating a final color transformed image by filtering the generated initial color transformed image using the processed noise amplification degrees and at least one of the raw image or luminance information of the raw image.

Abstract: A target tracking method and apparatus is provided. The target tracking apparatus includes a memory configured to store a neural network, and a processor configured to extract feature information of each of a target included in a target region in a first input image, a background included in the target region, and a searching region in a second input image, using the neural network, obtain similarity information of the target and the searching region and similarity information of the background and the searching region based on the extracted feature information, obtain a score matrix including activated feature values based on the obtained similarity information, and estimate a position of the target in the searching region from the score matrix.

Abstract: Disclosed are target tracking methods and apparatuses. The target tracking apparatus performs target tracking on an input image obtained in a first time period within a single time frame, using a light neural network in a second time period of the single time frame. The target tracking apparatus may perform target tracking on input images generated within the same time frame.

Abstract: A processor-implemented imaging method includes: obtaining initial homography information between a plurality of tele images that covers a field of view (FOV) of a wide image; receiving a user input of zooming a partial region of the wide image from a screen on which the wide image is displayed; stitching tele images corresponding to the partial region using the initial homography information, based on whether a zoom level corresponding to the user input exceeds a maximum zoom level of the wide image; and rendering the stitched tele images and displaying an image obtained by the rendering on the screen.

Abstract: A method and apparatus for color space conversion are provided. An apparatus for converting a color space includes at least one processor configured to convert an original full image of an original color space to a temporary image of a target color space, estimate a color space mapping parameter between the original color space and the target color space, the color space mapping parameter corresponding to the original full image, obtain a residual vector of the target color space based on the temporary image and the color space mapping parameter, convert the residual vector to a residual image of the target color space, and obtain a target full image of the target color space by combining the residual image with the temporary image.

Abstract: Disclosed are target tracking methods and apparatuses. The target tracking apparatus performs target tracking on an input image obtained in a first time period within a single time frame, using a light neural network in a second time period of the single time frame. The target tracking apparatus may perform target tracking on input images generated within the same time frame.

Abstract: A target tracking method and apparatus is provided. The target tracking apparatus includes a memory configured to store a neural network, and a processor configured to extract feature information of each of a target included in a target region in a first input image, a background included in the target region, and a searching region in a second input image, using the neural network, obtain similarity information of the target and the searching region and similarity information of the background and the searching region based on the extracted feature information, obtain a score matrix including activated feature values based on the obtained similarity information, and estimate a position of the target in the searching region from the score matrix.

Abstract: A white balancing device and a method of driving the same are provided. The white balancing device includes a map generator and an automatic white balancing (AWB) block. The map generator is configured to generate an indoor-outdoor map comparing a first image and a second image generated by performing a wide dynamic range (WDR) algorithm on the first image. The AWB block is configured to generate a first white point with respect to an outdoor region and a second white point with respect to an indoor region using the indoor-outdoor map.

Abstract: An image sensor includes a pixel array, a micro lens array and polarizing filters. The pixel array includes a plurality of pixels. The micro lens array is disposed on the pixel array and includes a plurality of micro lenses. The micro lens overlaps a pixel group of the pixel array. The polarizing filter is disposed on the micro lens and includes a plurality of polarizing filters having at least two different polarizing phases. Accordingly, the accuracy and the reliability of the local auto white balancing may be improved.