Canon Kabushiki Kaisha

Andrew Peter Bradley

James Philip Andrew

Kai Huang

Dominic Yip

Alison Joan Lennon

Andrew James Dorrell

Stephen James Hardy

Axel Becker

David Dominique M Poulouin

Edwin Ho

A method of compressing a current image of a sequence of images is disclosed. Firstly, the current image is transformed with a predetermined transform such as the DWT to provide a set of transform coefficients (step). The method then retrieves (step), for at least one transform coefficient of a current image, a predetermined number of bits, preferably two, of a corresponding transform coefficient of a previously compressed image of the sequence. The corresponding transform coefficient is truncated at a truncation bitplane and the retrieved bits are the least significant bits of the truncated corresponding transform coefficient. The transform coefficient of the current image is set to a new value that is a function of the retrieved bits (step) and bits of the transform coefficients of the current image are stored for use in compressing one or more subsequent images of the sequence (step).

Digital video compression

A method and apparatus for the detection of faces in a digital image captured under a variety of lighting conditions are provided. Rather than subjecting an entire image to computationally intensive face detection analysis, the image is subjected to computationally simply analysis to identify candidate pixels likely to be of skin colour. Only those pixels having such color are subject to computationally intensive face detection analysis. The pixels are selected by first selecting a color distribution model based on an image capture condition provided with the color digial image. Then, the color of the pixels is tested using the distribution model to determine those pixels having predominantly skin color. Preferably, the distribution model is selected from among a number of different distribution models with each of the distribution models associated with a different set of lighting conditions. In this way, a face detection process may accommodate a variety of lighting conditions associated with the capture of an image using a computationally simple analysis.

Face detection in digital images

Disclosed is a method and apparatus for clamping the output values of filtered image data comprising a mapping of discrete sample values are disclosed. For each discrete sample value of the mapping a maximum sample value and a minimum sample value of a plurality of input discrete samples values used to calculate the discrete sample value, is determined. The output value of the discrete sample value is clamped to the domain of the plurality of input discrete sample values utilising the maximum sample value and the minimum sample value, wherein the output value is dependent on a plurality of attributes of the plurality of input discrete sample values.

Color clamping

A method of interpolating image data is disclosed. The method accesses a first set of discrete sample values of the image data and calculates kernel values for each of the discrete sample values using one of a plurality of kernels. The kernel is selected depending upon an edge orientation indicator, an edge strength indicator, and an edge context indicator for each of the discrete sample values. The calculated kernel values are convolved with the discrete sample values to provide a second set of discrete sample values.

Method of kernel selection for image interpolation

A method and apparatus for determining edge information for image data comprising a mapping of discrete sample values are disclosed. An edge strength value is calculated for each of the discrete sample values of the image data to identify edge sample values, and an angle of orientation is stored for each of said edge sample values. The edge sample values and the stored angle of orientation for each of said discrete sample values are manipulated using a morphological process.

Method for kernel selection for image interpolation

A method of providing a representation of image data is disclosed. The method accesses a plurality of discrete sample values of the image data and determines an angle of orientation of the sample values. Kernel values for each of the discrete sample values are then calculated using one of a plurality of first kernels, depending upon the angle of orientation of the discrete sample value. The first kernels are constructed by rotating a second kernel from a first angle of orientation to a second angle of orientation. Finally, the kernel values are convolved with the discrete sample values to provide a representation of said image data.

Steerable kernel for image interpolation

A method of resolution conversion is disclosed. The method calculates a scaling factor for a first set of discrete data values compared to said second set of discrete data values. A plurality of filter function coefficients are calculated, based on the scaling factor, utilising at least one of a plurality of filter functions. The filter function coefficients are stored and later used to perform interpolation on the first set of discrete data values to generate the second set of discrete data values at a different resolution.

Continuous kernel image interpolation

A method of providing a representation of image data is disclosed. The method accesses a plurality of discrete sample values of the image data and calculates kernel values for each of the discrete sample values using a scaled kernel. The scaled kernel is constructed by transforming a kernel from a first range to a second range. In order to provide a representation of the image data, the kernel values are convolved with the discrete sample values.

Modified kernel for image interpolation

The method encodes a digital image to provide a compressed coded representation of the image. The method firstly performs a multi-level 2-D DWT transform ( ) on the entire image, which is arranged in a hierarchical order ( ). The sub-bands of the transform are then tiled ( ) into a number of blocks. The method then embed-bitplane encodes ( ) each block to visually lossless point. Afterwards, each encoded block is terminated ( ) at a bitplane that minimizes image distortion based on determined distortion measures and a desired total of determined block rates. Finally, the method concatenates ( ) the said terminated encoded blocks to form the coded representation.

Method and apparatus for representing a digital image to provide a coded representation

A method of creating a wavelet decomposition of a collection of data values is disclosed, the method including the steps of performing an initial decomposition of the data values into a series of components having low frequency components, high frequency components and components having a mixture of high and low frequencies; determining a first number of coefficients of each of the components having a magnitude exceeding a predetermined component threshold; creating a decomposition of each of the components into a series of sub components having low frequency sub-components, high frequency sub-components and sub-components having a mixture of high and low frequencies; determining a second number of coefficients of each of the sub-components having a magnitude exceeding a predetermined sub-component threshold; utilizing the first number and the second number to determine if the component should be decomposed into sub-components; and where the decomposition proceeds, applying the previous steps to each of the sub-components to a predetermined level of decomposition.

Method for digital data compression

A method is disclosed of encoding data including transforming the data into a series of components having different frequency characteristics, said components each including a series of coefficients; encoding said coefficients in a compact form, said coefficients including a binary string and said encoding including providing a first efficient encoding technique to encode the number of leading zeros in said binary string, and rounding said coefficients such that a rounded coefficient does not alter significantly the efficient coding of the leading zeros. Preferably, the rounding can includes discarding a first number of insignificant coefficients and adding one to the more significant coefficients only where the more significant coefficients are not zero. The invention has particular application when utilized in encoding wavelet transformed data and when utilized on image data.

Method for data compression

A method of producing a transform decomposition of data having an odd length the method comprising the steps of dividing the data into a portion having an even length of one element; performing a discrete wavelet transform on the even length data to produce low frequency subband data and high frequency subband data; adding the difference of the one element and an adjacent element to high frequency subband data. Preferably the transform is a Discrete Wavelet Transform utilised in the compression of image data.

Growing community of inventors

Castlecrag, Australia

Andrew Peter Bradley