Abstract: A vertical color detector group according to the present invention is formed on a semiconductor substrate and includes layers for collecting photons of different wavelength bands. The color detector group can be programmed to perform dynamic switching between sub-sampled color data and full measured color readout. The color detector group can also be configured in a portion of an array to emulate color filter array patterns, and programmed to dynamically alter the degree to which color information is sub-sampled. The programmable color detector groups can allow for switching between different levels of quality and resolution, allowing for selection of an optimal pattern based on image content or lighting conditions. By combining the color detector group of the present invention with conventional color filters, color filter arrays of more than three colors can be constructed.

Abstract: A method for storing a full Red, Green, Blue (RGB) data set. A full RGB data set is three-color image data captured with an imager array formed on a semiconductor substrate and comprising a plurality of vertical-color-filter detector groups. Each of the vertical color detector groups comprises three detector layers each configured to collect photo-generated carriers of a first polarity, separated by intervening reference layers configured to collect and conduct away photo-generated carriers of opposite polarity, the three detector layers being disposed substantially in vertical alignment with respect to one another and having different spectral sensitivities. The three-color image data is then stored as digital data in a digital storage device without performing interpolation on the three-color image data.

Abstract: A neural network has reduced requirements for storing intermodal weight values, as a result of a dual-precision training process. In the forward propagation of training samples, low-resolution weight values are employed. During back-propagation of errors to train the network, higher-resolution values are used. After training, only the lower resolution values need to be stored for further run-time operation, thereby reducing memory requirements.

Abstract: A method for controlling the exposure of an active pixel array electronic still camera includes the steps of: integrating photocurrent in each pixel during an integration time period; collecting overflow charge from all pixels in the array during the integration time period; developing an overflow signal as a function of the overflow charge; and terminating the integration time period when the overflow signal exceeds a preset threshold level selected to represent a desired reference exposure level.

Abstract: An integrated active pixel sensor array comprises a plurality of row select lines, each of said row select lines coupled to a source of a row-select signal; a plurality of source-follower drain row lines, each of said source-follower drain row lines coupled to a source of a source-follower drain row signal; a plurality of column output lines; a reset line coupled to a source of a reset signal; a source of reset potential; and a plurality of active pixel sensors, each pixel sensor associated with one row and one column of the array and including a photodiode having a first terminal coupled to a first reference potential and a second terminal, a Reset transistor having a gate coupled to the reset line, a drain coupled to the reset potential to reverse bias the photodiode, and a source coupled to the second terminal of the photodiode, a Source-Follower transistor having a gate coupled to the second terminal of the photodiode, a drain connected to the one of the plurality of source-follower drain row lines with w

Abstract: A storage pixel sensor disposed on a semiconductor substrate comprises a capacitive storage element having a first terminal connected to a fixed potential and a second terminal. A photodiode has an anode connected to a first potential and a cathode. A semiconductor reset switch has a first terminal connected to the cathode and a second terminal connected to a reset potential. A semiconductor transfer switch has a first terminal connected to the cathode and a second terminal connected to the second terminal of the capacitive storage element. A semiconductor amplifier has an input connected to the capacitive storage element and an output. The semiconductor reset switch and the semiconductor transfer switch each have a control element connected to a control circuit for selectively activating the semiconductor reset switch and the semiconductor transfer switch.

Abstract: A method for controlling the exposure of an active pixel array electronic still camera includes the steps of: integrating photocurrent in each pixel during an integration time period; collecting overflow charge from all pixels in the array during the integration time period; developing an overflow signal as a function of the overflow charge; and terminating the integration time period when the overflow signal exceeds a preset threshold level selected to represent a desired reference exposure level.

Abstract: A scanning circuit for use with an active pixel sensor array comprises a row-address generator configured to start at a selected row-start address, stop at a selected row-stop address, and increment row addresses by a factor K. A column-address generator is configured to start at a selected column-start address, stop at a selected column stop address, and increment column addresses by a factor K. Circuitry is coupled to the row address generator and the column address generator, for storing the row-start address, the row-stop address, the column-start address, the column-stop address and the factor K. A row decoder is coupled to the row-address generator and a column selector is coupled to the column-address generator. A plurality of row select lines are coupled to the row decoder, each one of the row select lines associated with a different row in the active pixel sensor array.

Abstract: An electronic camera system includes a lens system including at least one lens. A semiconductor sensor array having a plurality of pixels is optically coupled to the lens system. Each pixel generates an output signal that is a function of incident light. A sensor control circuit is adapted to produce sensor control signals for controlling the operation of the pixels in the semiconductor sensor array in response to user input. Circuitry is provided for producing from the semiconductor sensor array a first set of image output signals indicative of the intensity of the light at a first set of the pixels when the sensor control signals are in a first state, and a second set of image output signals indicative of the intensity of the light at a second set of the pixels when the sensor control signals are in a second state, the first set of pixels including more pixels than the second set of pixels.

Abstract: A color-separation prism assembly comprises first, second and third prisms. The first and second prisms have entrance surfaces, exit surfaces, and partially-reflecting surfaces, wherein the entrance surface of the second prism is separated by an air gap from the partially-reflecting surface of the first prism. The third prism has an entrance surface and an exit surface, wherein the entrance surface of the third prism is adjacent to the partially-reflecting surface of the second prism. The first prism has a cut-out serving as a flare-stop stop, and providing relief so as to allow the entrance surface of the third prism to slide across a plane disposed over the cut-out into a volume defined by the cut-out. The plane is formed as an extension of the partially-reflecting surface of the first prism. An optical axis passes through the first, second and third prisms. The optical axis passes at normal angles through the entrance surface of the first prism, and the exit surfaces of the first, second and third prisms.

Abstract: In a first embodiment an active pixel sensor includes a photodiode for capturing photocharge, a reset transistor for resetting the photodiode to a reset potential, and a readout transistor, and in a second embodiment an active pixel sensor includes a photodiode for capturing photocharge, a reset transistor for resetting the photodiode to a reset potential, a transfer transistor for transferring captured photocharge, and a readout transistor. In both embodiments, the readout transistor has a drain that is coupled to a first supply voltage during integration of photocharge and a second supply voltage during readout of the photocharge. Accordingly, the sensitivity of an active pixel sensor is increased by increasing the fill factor, the noise an active pixel sensor is reduced by increasing the relative size of the readout transistor, and the gain is compressive as the relative light intensity in an active pixel sensor increases.

Abstract: In a first embodiment an active pixel sensor includes a photodiode for capturing photocharge, a reset transistor for resetting the photodiode to a reset potential, and a readout transistor, and in a second embodiment an active pixel sensor includes a photodiode for capturing photocharge, a reset transistor for resetting the photodiode to a reset potential, a transfer transistor for transferring captured photocharge, and a readout transistor. In both embodiments, the readout transistor has a drain that is coupled to a first supply voltage during integration of photocharge and a second supply voltage during readout of the photocharge. Accordingly, the sensitivity of an active pixel sensor is increased by increasing the fill factor, the noise an active pixel sensor is reduced by increasing the relative size of the readout transistor, and the gain is compressive as the relative light intensity in an active pixel sensor increases.

Abstract: A color separating prism is disclosed for use in an electronic imaging systems such as a video or digital still-image camera. The prism separates an incoming light beam into red, green and blue light components and directs the separated light components onto adjacent imaging sensors. Beam-splitting interfaces of the prism are optically configured to admit approximately ten to twenty percent of a violet light contained in the incoming light beam into the red color channel. The prism may beneficially be optically coupled to a light-rejecting filter or mirror which rejects undesired far-red, far-violet and blue-green components of the light beam. In this manner, the resultant red, green and blue channels approximate a set of substantially non-negative color matching functions to facilitate highly colorimetrically accurate color imaging and thereby reduce or eliminate the need for post-imaging color correction.

Abstract: A storage pixel sensor comprises a photosensor selectively connectable to a reset potential; a switched buffer amplifier having a control terminal coupled to said photosensor, a first terminal connected to a source of a transfer signal, and a second terminal; a storage capacitor coupled to said second terminal of said switched buffer amplifier; and an amplifier coupled to said storage capacitor.

Abstract: A statistical classifier utilizes combined output values to determine posterior probabilities for certain output classes. In the field of handwriting recognition, compound characters are factored into classes of base letter forms and diacritical marks. A separate output activation value is produced for each base letter form and each diacritical mark. Pairs of output values, comprised of one value for a base letter form and one value for a diacritical mark, are combined to produce a posterior probability for every possible compound character, without requiring a network output for each possible class.

Abstract: To prevent overfitting a neural network to a finite set of training samples, random distortions are dynamically applied to the samples each time they are applied to the network during a training session. A plurality of different types of distortions can be applied, which are randomly selected each time a sample is applied to the network. Alternatively, a combination of two or more types of distortion can be applied each time, with the amount of distortion being randomly varied for each type.

Abstract: A statistical classifier for pattern recognition, such as a neural network, produces a plurality of output signals corresponding to the probabilities that a given input pattern belongs in respective classes. The classifier is trained in a manner such that low probabilities which pertain to classes of interest are not suppressed too greatly. This is achieved by modifying the amount by which error signals, corresponding to classes which are incorrectly identified, are employed in the training process, relative to error signals corresponding to the correct class. As a result, output probabilities for incorrect classes are not forced to a low value as much as probabilities for correct classes are raised.

Abstract: A statistical classifier that can be used for pattern recognition is trained to recognize negative, or improper patterns as well as proper patterns that are positively associated with desired output classes. A set of training samples includes both the negative and positive patterns, and target output values for the negative patterns are set so that no recognized class is indicated. The negative patterns are selected for training with less frequency than the positive patterns, and their effect on training is also modified, so that training is focused more heavily on improper patterns.

Abstract: A statistical classifier is trained in a manner to remove biasing due to unequal frequencies of unigram priors. The relative frequencies of all classes in a training set of sample patterns is determined. Training patterns are then selected from the set and skipped or repeated in dependence upon the relative frequency of the class to which they belong. In this manner, the presentation of samples is balanced across the classes.

Abstract: A method and system for approximating a Phong shading calculation for 3D renderings of realistic graphic images. The new method uses only a modest number of multiplies and adds to approximate a calculation that required divides, square roots and powers. The approximation uses approximate normalization, vector differences, and a shape function to simplify the processing and to improve performance significantly while still generating a graphic rendering that is very realistic.