Motion vector estimating apparatus

• US 5,068,722 A
• Filed: 11/16/1990
• Issued: 11/26/1991
• Est. Priority Date: 01/23/1990
• Status: Expired due to Term

• Current Assignee: Victor Company of Japan Limited (JVC Kenwood Corporation)
• Original Assignee: Victor Company of Japan Limited (JVC Kenwood Corporation)
• Inventors: Sugiyama, Kenji
• Primary Examiner(s): Peng, John K.
• Application Number:
  US07/614,478
• Time in Patent Office: 375 Days
• Field of Search: 358/105, 358/135, 358/136, 358/140
• US Class Current: 348/420.1
• CPC Class Codes: G06T 7/231 using full search
• CPC Class Codes: H04N 19/51 Motion estimation or motion compensation
• CPC Class Codes: H04N 5/145 Movement estimation for video coding H04N19/51

Abstract

A motion vector estimating apparatus includes a plurality of block matching estimators generating block matching values from an input video signal respectively. The block matching values represent likelihoods of motion vectors with respect to an actual motion respectively. Each of the block matching values is periodically reset. A plurality of convolution adders receive the block matching values respectively. Each of the convolution adders provides predetermined weights to the block matching values and adds the weighted block matching values to obtain an output value. The smallest of the output values from the convolution adders is selected and a vector related to the smallest values is outputted as a final motion vector.

Claims

1. A motion vector estimating apparatus comprising:

a plurality of block matching estimators generating block matching values from an input video signal respectively, wherein the block matching values represent likelihoods of motion vectors with respect to an actual motion respectively;

a plurality of convolution adders receiving the block matching values respectively, wherein each of the convolution adders provides predetermined weights to the block matching values and adds the weighted block matching values to obtain an output value; and

means for selecting a smallest value of output values from the convolution adders, and outputting a vector related to the smallest value as a final motion vector.

2. A motion vector estimating apparatus as recited in claim 1 wherein each of said block matching estimators includes means for periodically resetting the block matching values.

3. A motion vector estimating apparatus as recited in claim 2 wherein said means for periodically resetting the block matching values comprises an adder and a block reset signal applied thereto.

4. A motion vector estimating apparatus comprising:

a plurality of block matching estimators generating block matching values from an input video signal respectively, wherein the block matching values represent likelihoods of motion vectors with respect to an actual motion respectively;

a plurality of convolution adders receiving the block matching values respectively, wherein each of the convolution adders includes nonlinear converters, and the converters convert block matching values for blocks around an estimation-object block with a nonlinear conversion characteristic, and wherein each of the convolution adders provides predetermined weights to output values from the nonlinear converters and the block matching value for the estimation-object block and adds the weighted values to obtain an output value; and

means for selecting a smallest value of output values from the convolution adders, and outputting a vector related to the smallest as a final motion vector.

5. The motion vector estimating apparatus of claim 4 wherein each of the convolution adders comprises means for changing the nonlinear conversion characteristic in response to the block matching value for the estimation-object block.

6. A motion vector estimating apparatus as recited in claim 4 wherein each of said block matching estimators includes means for periodically resetting the block matching values.

7. A motion vector estimating apparatus as recited in claim 6 wherein said means for periodically resetting the block matching values comprises an adder and a block reset signal applied thereto.

8. An apparatus for processing a video signal in which a frame of the video signal is divided into a predetermined number N1 of blocks, and the video signal is processed sequentially block by block, the apparatus comprising:

a predetermined number N2 of block matching estimators generating respective block matching values for a block as functions of an input video signal, wherein the block matching values represent likelihoods of motion vectors with respect to an actual motion respectively;

N2 convolution adders respectively receiving the block matching values from the block matching estimators, wherein each of the convolution adders provides a predetermined number N3 of predetermined weights to respective block matching values which are sequentially outputted from the respective block matching estimator and which correspond to adjacent blocks, and adds the weighted block matching values to obtain an output value for a current object block; and

means for selecting a smallest value of output values from the convolution adders, and outputting a vector related to the smallest value as a final motion vector for the current object block.

9. An apparatus for processing a video signal in which a frame of the video signal is divided into a predetermined number N1 of blocks, and the video signal is processed sequentially block by block, the apparatus comprising:

a predetermined number N2 of block matching estimators generating respective block matching values for a block as functions of an input video signal, wherein the block matching values represent likelihoods of motion vectors with respect to an actual motion respectively;

N2 convolution adders respectively receiving the block matching values from the block matching estimators, wherein each of the convolution adders includes nonlinear converters, and the converters convert block matching values, which are sequentially outputted from the respective block matching estimator and which correspond to blocks around a current object block, with a nonlinear conversion characteristic, and wherein each of the convolution adders provides predetermined weights to output values from the nonlinear converters and the block matching value for the current object block and adds the weighted values to obtain an output value for a current object block; and

means for selecting a smallest value of output values from the convolution adders, and outputting a vector related to the smallest value as a final motion vector for the current object block.