IMAGE PROCESSING CIRCUIT AND IMAGE DETECTION DEVICE

• US 20140294293A1
• Filed: 03/19/2014
• Published: 10/02/2014
• Est. Priority Date: 03/29/2013
• Status: Active Grant

• Current Assignee: Megachips Corporation
• Original Assignee: Megachips Corporation
• Inventors: YAMAMURA, Naotsugu
• Granted Patent: US 9,053,387 B2
• Time in Patent Office: Days
• US Class Current: 382/159
• CPC Class Codes: G06F 18/214 Generating training patterns; Bootstrap methods, e.g. bagging or boosting
• CPC Class Codes: G06T 2207/30201 Face
• CPC Class Codes: G06V 10/467 Encoded features or binary features, e.g. local binary patterns [LBP]
• CPC Class Codes: G06V 40/161 Detection; Localisation; Normalisation
• CPC Class Codes: G06V 40/171 Local features and components; Facial parts eye characteristics G06V40/18; Occluding parts, e.g. glasses; Geometrical relationships

Abstract

Each second selection circuit selects, out of a plurality of evaluation values, an evaluation value being in a predetermined relative positional relation with a first evaluation value as an evaluation value outputted from a first selection circuit, and outputs the selected value. The predetermined relative positional relations are different from one another among a plurality of second selection circuits. Every time a second evaluation value is outputted from the second selection circuit corresponding to the integration circuit, the integration circuit reads a weigh value corresponding to a combination of the second evaluation value and the first evaluation, which makes a pair with the second evaluation and is outputted from the first selection circuit, from a storage circuit corresponding to the second selection circuit and integrates the read values. An addition circuit at least adds a plurality of integrated values outputted from a plurality of integration circuits, and an addition value obtained thereby becomes a probability value.

Claims

1. An image processing circuit for obtaining a probability value indicative of a probability that an input image is a specific image, said circuit comprising:

a first selection circuit for selecting one evaluation value out of a plurality of evaluation values which indicate respective characteristics of a plurality of pixels included in said input image and are allocated with positions different from one another, and outputting the selected value;

a plurality of second selection circuits, each for selecting one evaluation value from said plurality of evaluation values and outputting the selected value;

a plurality of storage circuits provided respectively corresponding to said plurality of second selection circuits;

a plurality of integration circuits provided respectively corresponding to said plurality of second selection circuits;

an addition circuit; and

a control circuit for controlling selection of the evaluation values in said first selection circuit and said plurality of second selection circuits,

wherein

said first selection circuit sequentially changes one evaluation value selected out of said plurality of evaluation values, to change an evaluation value to be outputted,

each of said plurality of second selection circuits selects, out of said plurality of evaluation values, one evaluation value being in a predetermined relative positional relation with a first evaluation value as an evaluation value which is outputted from said first selection circuit, and outputs the selected value,

said predetermined relative positional relations are different from one another among said plurality of second selection circuits,

each of said plurality of storage circuits stores a weigh value based on a learning sample concerning said specific image in accordance with a combination of a second evaluation value as an evaluation value which is outputted from the corresponding second selection circuit and said first evaluation value which makes a pair with the second evaluation value and is outputted from said first selection circuit,

every time said second evaluation value is outputted from the second selection circuit corresponding to the integration circuit, the integration circuit reads, from the storage circuit corresponding to the second selection circuit, said weigh value in accordance with a combination of the second evaluation value and said first evaluation value which makes a pair with the second evaluation value and is outputted from said first selection circuit, to integrate the read values and outputs the consequently obtained integrated value, and

said addition circuit at least adds a plurality of integrated values outputted from said plurality of integration circuits, so that an addition value is obtained and becomes said probability value.

2. The image processing circuit according to claim 1, wherein said weigh values stored in said plurality of storage circuits are rewritable.

3. The image processing circuit according to claim 1, wherein

said image processing circuit is capable of obtaining said probability value concerning each of plural kinds of specific images, the probability value being indicative of a probability that said input image is the specific image,

each of said plurality of storage circuits stores plural kinds of weigh values which respectively correspond to said plural kinds of specific images, each of said plural kinds of weigh values being based on a learning sample concerning the corresponding specific image in accordance with a combination of said second evaluation value outputted from the second selection circuit corresponding to the storage circuit and said first evaluation value which makes a pair with the second evaluation value and is outputted from said first selection circuit,

said control circuit sets the kind of weigh value to be used with respect to said plurality of integration circuits, and

every time said second evaluation value is outputted from the second selection circuit corresponding to the integration circuit, the integration circuit reads, from the storage circuit corresponding to the second selection circuit, the weigh value of the kind set by the control circuit in accordance with a combination of the second evaluation value and said first evaluation value which makes a pair with the second evaluation value and is outputted from said first selection circuit, to integrate the read values.

4. The image processing circuit according to claim 2, wherein said control circuit further includes a register which stores setting information for setting an operation to select the evaluation value in each of said plurality of second selection circuits.

5. The image processing circuit according to claim 1, wherein

said image processing circuit is capable of obtaining a second probability value indicative of a probability that said input image is a second specific image of which a kind is different from that of a first specific image as said specific image,

said plurality of second selection circuits have, as operation modes, a first mode which becomes effective at the time of obtaining a first probability value as said probability value concerning said first specific image, and a second mode which becomes effective at the time of obtaining said second probability value,

said control circuit sets the operation mode of said plurality of second selection circuits,

a position of said second evaluation value in said first mode and a position of said second evaluation value in said second mode, both making a pair with said same first evaluation value, are horizontally symmetrical or vertically symmetrical each other with respect to a position of the first evaluation value in each of said plurality of second selection circuits,

in a case where the second selection circuit corresponding to the integration circuit is on said first mode, every time said second evaluation value is outputted from the second selection circuit, the integration circuit reads from the storage circuit corresponding to the second selection circuit said weigh value in accordance with a combination of the second evaluation value and said first evaluation value which makes a pair with the second evaluation value and is outputted from said first selection circuit, to integrate the read values and outputs the consequently obtained integrated value as a first integrated value,

while in a case where the second selection circuit corresponding to the integration circuit is on said second mode, every time said second evaluation value is outputted from the second selection circuit, the integration circuit reads from the storage circuit corresponding to the second selection circuit said weigh value in accordance with a combination of the second evaluation value and said first evaluation value which makes a pair with the second evaluation value and is outputted from said first selection circuit, to integrate the read values and outputs the consequently obtained integrated value as a second integrated value,

said addition circuit at least adds a plurality of integrated values outputted from said plurality of integration circuits in a case where said plurality of second selection circuits are on said first mode, so that a first addition value is obtained and becomes said first probability value, and

said addition circuit at least adds a plurality of integrated values outputted from said plurality of integration circuits in a case where said plurality of second selection circuits are on said second mode, so that a second addition value is obtained and becomes said second probability value.

6. An image detection device, comprising:

an image processing circuit for taking, as an input image, each of a plurality of images selected from a processing object image and obtaining a probability value indicative of a probability that the input image is a specific image; and

a specification section for specifying said specific image in said processing object image based on said probability value found in said image processing circuit,

wherein

said image processing circuit has;

a first selection circuit for selecting one evaluation value out of a plurality of evaluation values which indicate respective characteristics of a plurality of pixels included in said input image and are allocated with positions different from one another, and outputting the selected value;

a plurality of second selection circuits, each for selecting one evaluation value from said plurality of evaluation values and outputting the selected value;

a plurality of storage circuits provided respectively corresponding to said plurality of second selection circuits;

a plurality of integration circuits provided respectively corresponding to said plurality of second selection circuits;

an addition circuit; and

a control circuit for controlling selection of the evaluation values in said first selection circuit and said plurality of second selection circuits,

said first selection circuit sequentially changes one evaluation value selected out of said plurality of evaluation values, to change an evaluation value to be outputted,

each of said plurality of second selection circuits selects, out of said plurality of evaluation values, one evaluation value being in a predetermined relative positional relation with a first evaluation value as an evaluation value which is outputted from said first selection circuit, and outputs the selected value,

said predetermined relative positional relations are different from one another among said plurality of second selection circuits,

each of said plurality of storage circuits stores a weigh value based on a learning sample concerning said specific image in accordance with a combination of a second evaluation value as an evaluation value which is outputted from the corresponding second selection circuit and said first evaluation value which makes a pair with the second evaluation value and is outputted from said first selection circuit,

every time said second evaluation value is outputted from the second selection circuit corresponding to each of said plurality of integration circuits, the integration circuit reads, from the storage circuit corresponding to the second selection circuit, said weigh value in accordance with a combination of the second evaluation value and said first evaluation value which makes a pair with the second evaluation value and is outputted from said first selection circuit, to integrate the read values and outputs the consequently obtained integrated value, and

said addition circuit at least adds a plurality of integrated values outputted from said plurality of integration circuits, so that an addition value is obtained and becomes said probability value.

7. The image detection device according to claim 6, wherein

said image processing circuit obtains said probability value concerning each of plural kinds of specific images, the probability value being indicative of a probability that said input image is the specific image,

each of said plurality of storage circuits stores a plural kinds of weigh values which respectively correspond to said plural kinds of specific images, each of said plural kinds of weigh value being based on a learning sample concerning the corresponding specific image in accordance with a combination of said second evaluation value outputted from the second selection circuit corresponding to the storage circuit and said first evaluation value which makes a pair with the second evaluation value and is outputted from said first selection circuit,

said control circuit sets the kind of weigh value to be used with respect to said plurality of integration circuits,

every time said second evaluation value is outputted from the second selection circuit corresponding to the integration circuit, the integration circuit reads, from the storage circuit corresponding to the second selection circuit, the weigh value of the kind set by the control circuit in accordance with a combination of the second evaluation value and said first evaluation value which makes a pair with the second evaluation value and is outputted from said first selection circuit, to integrate the read values, and

said specification section specifies a predetermined entire image with each of said plural kinds of specific images taken as a part thereof in said processing object image based on said probability values concerning said plural kinds of specific images obtained in said image processing circuit.

8. The image detection device according to claim 7, wherein

said plural kinds of specific images are images of a plurality of parts constituting a human body, and

said predetermined entire image is an image of a human whole body.

9. The image detection device according to claim 6, wherein

said image processing circuit obtains a second probability value indicative of a probability that said input image is a second specific image of which a kind is different from that of a first specific image as said specific image,

said plurality of second selection circuits have, as operation modes, a first mode which becomes effective at the time of obtaining a first probability value as said probability value concerning said first specific image, and a second mode which becomes effective at the time of obtaining said second probability value,

said control circuit sets the operation mode of said plurality of second selection circuits,

a position of said second evaluation value in said first mode and a position of said second evaluation value, both making a pair with said same first evaluation value, are horizontally symmetrical or vertically symmetrical each other with respect to a position of the first evaluation value in each of said plurality of second selection circuits,

in a case where the second selection circuit corresponding to the integration circuit is on said first mode, every time said second evaluation value is outputted from the second selection circuit, the integration circuit reads from the storage circuit corresponding to the second selection circuit said weigh value in accordance with a combination of the second evaluation value and said first evaluation value which makes a pair with the second evaluation value and is outputted from said first selection circuit, to integrate the read values and outputs the consequently obtained integrated value,

while in a case where the second selection circuit corresponding to the integration circuit is on said second mode, every time said second evaluation value is outputted from the second selection circuit, the integration circuit reads from the storage circuit corresponding to the second selection circuit said weigh value in accordance with a combination of the second evaluation value and said first evaluation value which makes a pair with the second evaluation value and is outputted from said first selection circuit, to integrate the read values and outputs the consequently obtained integrated value,

said addition circuit at least adds a plurality of integrated values outputted from said plurality of integration circuits in a case where said plurality of second selection circuits are on said first mode, so that a first addition value is obtained and becomes said first probability value,

said addition circuit at least adds a plurality of integrated values outputted from said plurality of integration circuits in a case where said plurality of second selection circuits are on said second mode, so that a second addition value is obtained and becomes said second probability value, and

said specification section specifies a predetermined entire image with said first and second specific images taken as a left-side part and a right-side part or an upper-side part and a lower-side part in said processing object image based on said first and second probability values obtained in said image processing circuit.

10. The image detection device according to claim 9, wherein

said first and second specific images are images of a left-side part and a right-side part of a human face, and

said predetermined entire image is an image of an entire human face.