With known methods for predictive decorrelation this decorrelation only applies to one or more previously scanned elements of the same picture line. In accordance with the invention a picture to be scanned is divided into sub-pictures wherein a number of picture lines equal to the number of lines comprised in a sub-picture is intermediately stored in shift registers, a processing circuit being connected to the last stages of these shift registers, which processing circuit multiplies the values of the picture elements of one sub-picture with the same number of coefficients stored in the processing device and which adds the products. In a preferred embodiment the difference values obtained in this manner are also decorrelated in a second stage, the matrix of the difference values being divided into difference value sub-pictures which are shifted, preferably diagonally, relative to the original sub-pictures. A considerably larger surrounding area of each picture element is thus involved in the prediction.
A redundancy reduction coding method of a multi-tone picture element comprises the steps of selecting reference picture elements from picture elements already scanned and transmitted, selecting a single reference picture element as a base reference element and then evaluating a plurality of first tone level differences between the base reference picture element and the reference picture elements other than the base reference picture element and a second tone level difference between the base reference picture element and an objective picture element to be transmitted. Then, if and when all of the tone level differences are within a predetermined region, the second tone level difference is converted into a predictive order in which a second tone level difference appears for a particular combination of the plurality of first tone level differences, the predictive order being produced in advance based on statistics of a standard picture. Thereafter, the predictive order is converted into a predetermined assigned code. The code is a variable length code in which a shorter code is assigned to a higher rank of the predictive order.
An improved circuit for performing the S-transform and inverse S-transform on digital images. Data is fed back so that only one stage of adder/subtracters is required. The circuit elements have little idle time, making for efficient VLSI implementation.
Predictive errors are grouped into a plurality of quantizing ranges, and the predictive errors in the respective quantizing ranges are represented by representative quantizing values assigned respectively to the quantizing ranges. The predictive error data items in each of the quantizing ranges are processed to provide a weighted mean value based on the frequency distribution thereto. The weighted mean value or a value close thereto is used as a representative quantized value for that quantizing range.
A motion wavelet transform zero tree codec achieves high compression and is implemented in hardware of modest size and at very low cost. A wavelet transform is combined with a tree walk technique for encoding the resulting wavelet coefficients. A 2-6 wavelet transform is used. Wavelet coefficients from the transform are represented in a pyramid of wavelet coefficients. An array of zero trees are formed from the pyramid to hold the wavelet coefficients, one coefficient to each node. Significance values for each node are calculated to assist with the tree walk and encoding. Each zero tree is traversed to produce an output of encoded bits. Encoded bits are output directly during the tree walk.
A technique for compression and expansion of a function defined upon an M-dimensional manifold embedded in N-dimensional space uses a second generation wavelet transform and a modified zerotree bit-encoding scheme. Typically, a function is defined upon a two-dimensional manifold embedded in three-dimensional space, such as a sphere. A geometric base is chosen as a coarse initial model of the manifold. Second generation wavelets for the function are calculated using a triangular subdivision scheme in order to subdivide the geometric base in order to produce a refined triangular mesh. The wavelet coefficients are defined at the vertices of the triangles in the triangular mesh. A tree structure is created in which each node of the tree structure represents an associated triangle of the triangular mesh. Each triangle in the mesh is recursively subdivided into four subtriangles and each associated node in the tree structure also has four children, which correspond to the four subtriangles. Each wavelet coefficient defined at a particular vertex in the triangular mesh is uniquely assigned to a single one of the triangles at a next higher level of subdivision, such that each triangle at the next higher level of subdivision has from zero to three assigned wavelet coefficients. Using a modified zerotree encoding scheme, values of the wavelet coefficients are processed bit plane by bit plane, outputting bits indicative of significant nodes and their descendants. Sign bits and data bits are also output. An expansion technique inputs bits according to the modified zerotree scheme into the tree structure in order to define wavelet coefficients. An inverse second generation wavelet transform is used to synthesize the original function from the wavelet coefficients.
