US 5,479,218 A · Motion vector detecting apparatus

Abstract

By expanding the technique of detection of a motion vector by the of time-space differential method, (1) repetitive calculations are not needed, (2) different motion vectors near the boundary of an object can be detected, and (3) coupling of the region segmentation and the framework is enabled. For this, a horizontal direction differentiating filter (104), a vertical direction differentiating filter (105), and a time direction differentiating filter (106) calculate the time-space differential (.differential.I/.differential.x, .differential.I/.differential.y, .differential.I/.differential.t) in formula (2) necessary for motion vector estimation in every pixel. A feature vector combining part (110) generates a sample vector coupling the time-space differential value and position for a random image position (x, y), and clusters from pairs of samples and classes, to minimize the difference in a maximum likelihood class determining part (112) and a maximum likelihood class data changing part (113). The motion vector is obtained from the third eigen vector of covariance matrix of the time-space differential value.

Drawings

Drawings are temporarily unavailable. Open the PDF at USPTO

Drawings

Claims4

Dependent claims

A motion vector detecting apparatus comprising:

(a) a memory for holding video signals which are coded and composed of frame units,
(b) a horizontal differentiating filter for reading out luminance data from the memory and differentiating in a horizontal direction of arbitrary image coordinates,
(c) a vertical differentiating filter for reading out luminance data from the memory and differentiating in a vertical direction of the image coordinates,
(d) a time differentiating filter for reading out the luminance data of preceding and succeeding frames from the memory, and differentiating in a time direction,
(e) feature vector combining means for combining a luminance differential vector (Ix, Iy, It), including results from the horizontal differentiating filter, the vertical differentiating filter and the time differentiating filter, with a position vector (x, y) of the image coordinates, to obtain a sample vector,
(f) a memory for holding a plurality of estimations of a motion vector (u, v) and a position vector mean (x, y) of the sample vector as class data,
(g) nearest class determining means for determining a distance between the class data and the sample vector from a formula (x-x, y,-y) and a formula (u*Ix+v*Iy+It), and achieving a correspondence between the sample vector and the nearest class data, and
(h) class data changing means for changing the class data in a manner that a distance sum between the class data and one or more sample vectors becomes smaller, from one or more sample vectors and the class data of which correspondence has been achieved by the nearest class determining means.

5 more elements

Dependencies are inferred from the claim text.

Description3,343 words

Summary1,380 words

#1. Field of the Invention

#The present invention relates to a motion vector detecting apparatus for coding images, reducing noise, tracing objects, etc.

#2. Related Art of the Invention

Brief description of the drawings4 figures

#FIG. 1 is a structural diagram of a detecting apparatus for a motion vector in the prior art.

#FIG. 2 is a structural diagram of a first embodiment of the detecting apparatus of a motion vector.

#FIG. 3 is a structural diagram of a second embodiment of the detecting apparatus of a motion vector.

Detailed description1,881 words

#Referring to the drawings, some of the embodiments of the invention are described in detail below: The first embodiment of the invention is described by reference to FIGS. 2, 4a and 4b.

#FIG. 2 is a structural diagram of a detecting apparatus of a motion vector in the first embodiment of the invention. FIGS. 4a and 4b are diagrams showing the principle of estimation of motion in the first embodiment. In FIG. 2, numeral 101 is an A/D converter, 102, 103 are frame memories, 104 is a horizontal direction differentiating filter, 105 is a vertical direction differentiating filter, 106 is a time direction differentiating filter, 107 is a horizontal direction differential image memory, 108 is a vertical direction differential image memory, 109 is a time direction differential image memory, 110 is a feature vector combining part, 111 is a random address generating part, 112 is a maximum likelihood class determining part as an example of a nearest class determining means, 113 is a maximum likelihood class changing part, and 114 is a class data memory bank.

#The image entered through the A/D 101 is delayed by 1 frame time in the frame memory 102, and sequentially passes through the frame memory 103.

Facts

Application

08/141,051Filed 26 Oct 1993

Priority

JP 4-2941202 Nov 1992

Granted

26 Dec 1995

Term ended

26 Oct 2013

Estimated · how

Pre-1995 term; greater of 17 years from grant or 20 years from filing when eligible. Maintenance-fee lapses, terminal disclaimers and statutory extensions have not been checked. The continuity chain may be incomplete; this is not an enforceability determination.

Classification5 codes

H04N 19/51
Involving temporal prediction: motion estimation or motion compensation
G06T 7/269
Analysis of motion: using gradient-based methods
G06T 7/277
Analysis of motion: involving stochastic approaches, e.g. using Kalman filters
H04N
Pictorial communication

Motion vector detecting apparatus

Life of this patent

Motion vector detecting apparatus

Life of this patent

Life of this patent

Abstract

Drawings

Drawings

Claims4

Description3,343 words

Drawings

Classification5 codes

Life of this patent

Description3,343 words