Gil Algorithms projects:
Gil Algorithms projects
Aerophoto
  • Developed an algorithm to find common points in aerial photographs of varying landscapes, such as trees swaying in the wind.
  • Developed an algorithm for fast location of out-of-focus images, provided that only one image is available in each scene.
Image processing
  • Developed an algorithm for automatic correction of lens shading for CMOS image sensors that does not require information on the type of illumination.
  • Developed an algorithm for the transformation of Bayer into RGB, making it possible to restore 1 pixel size objects and to provide a low level of color aliasing.
  • Developed an intense color noise reduction algorithm saving the fine details of an image.
  • Developed and patented a system of automatic calibration of pixel crosstalk not requiring involvement of any special equipment.
  • Developed an algorithm for the alignment of four images of the same scene in such a way that an image with four times greater resolution is obtained; at the same time, the algorithm allows for the preservation of the sharpness and the naturalness of the edges.
  • Developed an algorithm for the high-quality re-creation of an image taken simultaneously by several cameras with different technical characteristics.
  • Developed an algorithm to process the image of a drop of sweat falling on a test plate to define its chemical composition.
  • Developed an algorithm to eliminate the pixel crosstalk effect with simultaneous: a) color recovery (the colors were corrected); b) recovery of image sharpness; c) decrease in noise factor (the algorithm application results in significantly lower noise increase than during application of standard algorithms for color correction and sharpening).
  • Developed a chain of image processing algorithms for a digital camera, including: shading correction, Gr/Gb correction, Bayer to RGB, denoising, anti-aliasing, sharpening, and color correction. Obtained a patent for the Bayer to RGB conversion algorithm.
  • Developed a method for integrating Extended Depth of Field technology into a usual standard chain of image processing in a camera. The integration does not increase the number of transistors employed and, as a consequence, the electronic circuit area remains unchanged.
  • Developed an algorithm allowing for the change of a picture in the Bayer format to avoid, even after its conversion into the RGB format by low quality algorithms, color noise or aliasing.
  • Prepared a survey of algorithms for improving the contrast range and dynamic depth of images.
  • Compared the characteristics of the chains of image algorithms used in modern cameras: the chains of algorithms of all leading manufacturers have been studied, and for each unit a detailed analysis has been prepared to indicate the algorithms used.
  • Developed a new algorithm correcting the white balance of an image, solving major problems that exist today, including: a) proper rendering of skin color, sky, and vegetation; b) establishment of light exposure in the event that the frame has no white or grey color objects; c) balance of colors in pictures taken in warm lighting.
Video processing
  • Developed an algorithm that allows for an instant, within-one-frame, change in the transfer rate of a MPEG-4 video stream, to adapt a video signal transmitted via cellular networks to the change in reception quality of a specific subscriber. Provided that the image is not torn, it either smoothly loses quality or is sharpened.
  • Developed an effective MPEG-4 codec that applies to coding television channels and video clips for transmission via a 3G cellular network. The algorithm includes automatic image segmentation, various frequency upgrades of foreground and background objects, comparison of various methods of video signal compression in real-time mode, and selection of the best compression method for each scene.
Dynamic control
  • Developed a system of low-level filters to substantially upgrade the user experience when working with the new touch screens and touch pads. The parameters of these filters change automatically depending on the type of operation that the user is performing.
  • Developed an algorithm to control the exposure, laser radiant intensity, and parameters of a digital camera for photographing complex contrast scenes in the infrared spectrum. Provided that iterated convergence is not used, in most cases the optimal parameters are chosen in one frame.
Medical equipment
  • Developed and programmed an algorithm for recognition of blood vessels and calculation of their characteristics (blood flow, vessel diameter, walls elasticity, etc.) by an ultrasonic picture in real time mode, 80 MB of data per second.
  • Developed an algorithm to remove noise in ultrasonic "echo" and "Doppler" images. The algorithm builds the most suitable noise model in real-time mode, since in the course of measurement the character and amplitude of noise varies, on average, several times per second.
Statistics
  • Developed an algorithm that finds interdependent values in a large data stream. Provided that a nonlinear dependence of the values observed is assumed, the random noise level can materially exceed the functional connection level.
    In practice, this method is applied in the analysis of exchange indexes, and also in biological and geological experiments.
Consumer electronics
  • Developed a mathematical model of resistive touch screens that allows simultaneous touch with several fingers.
Algorithms performance optimization
  • Bad pixel correction, pixel crosstalk correction, residual Gr/Gb correction and high-quality de-noising algorithms have been merged into one block, resulting in a hardware block that requires only six line buffers.
  • Optimized the algorithm that searches for the points for a three-dimensional scene reconstruction. After optimization, the performance of the algorithm increased 12 times without any loss of quality.