Mithun Das Gupta

US Patent:

8160380, Apr 17, 2012

Filed:

Jul 8, 2010

Appl. No.:

12/832757

Inventors:

Mithun Das Gupta - Cupertino CA, US
Jing Xiao - Cupertino CA, US

Assignee:

Seiko Epson Corporation - Tokyo

International Classification:

G06K 9/40
H04N 3/14

US Classification:

382260, 348272

Abstract:

Application of an image filtering algorithm, which defines an algorithm window within which a center pixel is processed relative to the other pixels within the algorithm window, is improved by use of an extended window larger than and encompassing the algorithm window. This approached is applied with an edge preserving filter that works on the principle of matting affinity and allows a better representation of the range filter term in bilateral class filters. An approximate bi-affinity filter whose output is shown to be very similar to the traditional bilateral filter is defined. The present technique has the added advantage that no color space changes are required and hence an input image can be handled in its original color space.

US Patent:

8194918, Jun 5, 2012

Filed:

Feb 5, 2010

Appl. No.:

12/701349

Inventors:

Mithun Das Gupta - Cupertino CA, US
Jing Xiao - Cupertino CA, US

Assignee:

Seiko Epson Corporation - Tokyo

International Classification:

G06K 9/00

US Classification:

382100

Abstract:

A watermarking system uses distinct bit patterns to identify a logic 0, a logic 1, and a marker bit, which demarcates segments of logic bit information. Marker bits, which are printed on both foreground and background areas of an image, outline message blocks. In message extraction, a preprocessing step removes any white boarders, identifies the best defined corner of a message block, crops the image, and rotates the image to place the identified corner at the top-left corner. Message extraction scans the rotated image in window segments of increasing size during multiple cycles. During each cycle, if a bit pattern cannot be identified as a data bit, then the size of the examined bit area is increased and rechecked to see it specifically is a marker bit. If no bit information can be definitively identified, then it is assigned a logic bit value based on a 50% random assignment.

US Patent:

8243985, Aug 14, 2012

Filed:

Feb 5, 2010

Appl. No.:

12/701290

Inventors:

Mithun Das Gupta - Cupertino CA, US
Jing Xiao - Cupertino CA, US

Assignee:

Seiko Epson Corporation - Tokyo

International Classification:

G06K 9/00
H04L 9/32

US Classification:

382100, 713176

Abstract:

A watermarking system uses distinct bit patterns to identify a logic 0, a logic 1, and a marker bit, which demarcates segments of logic bit information. Marker bits, which are printed on both foreground and background areas of an image, outline message blocks. In message extraction, a preprocessing step removes any white boarders, identifies the best defined corner of a message block, crops the image, and rotates the image to place the identified corner at the top-left corner. Message extraction scans the rotated image in window segments of increasing size during multiple cycles. During each cycle, if a bit pattern cannot be identified as a data bit, then the size of the examined bit area is increased and rechecked to see it specifically is a marker bit. If no bit information can be definitively identified, then it is assigned a logic bit value based on a 50% random assignment.

US Patent:

8285033, Oct 9, 2012

Filed:

Apr 1, 2010

Appl. No.:

12/752615

Inventors:

Mithun Das Gupta - Cupertino CA, US
Jing Xiao - Cupertino CA, US

Assignee:

Seiko Epson Corporation - Tokyo

International Classification:

G06K 9/00
H04N 3/14

US Classification:

382162, 348272

Abstract:

An edge preserving filter that works on the principle of matting affinity allows a better representation of the range filter term in bilateral class filters. The definition of the affinity term can be relaxed to suit different applications. An approximate bi-affinity filter whose output is shown to be very similar to the traditional bilateral filter is defined. The present technique has the added advantage that no color space changes are required and hence an input image can be handled in its original color space. This is a big benefit over the traditional bilateral filter, which needs conversion to perception based spaces, such as CIELAB, to generate results close to the present invention. The full bi-affinity filter preserves very minute details of the input image, and thus permits an enhanced zooming functionality.

US Patent:

8300882, Oct 30, 2012

Filed:

Feb 5, 2010

Appl. No.:

12/701311

Inventors:

Mithun Das Gupta - Cupertino CA, US
Jing Xiao - Cupertino CA, US

Assignee:

Seiko Epson Corporation - Tokyo

International Classification:

G06K 9/00

US Classification:

382100

Abstract:

A watermarking system uses distinct bit patterns to identify a logic 0, a logic 1, and a marker bit, which demarcates segments of logic bit information. Marker bits, which are printed on both foreground and background areas of an image, outline message blocks. In message extraction, a preprocessing step removes any white boarders, identifies the best defined corner of a message block, crops the image, and rotates the image to place the identified corner at the top-left corner. Message extraction scans the rotated image in window segments of increasing size during multiple cycles. During each cycle, if a bit pattern cannot be identified as a data bit, then the size of the examined bit area is increased and rechecked to see it specifically is a marker bit. If no bit information can be definitively identified, then it is assigned a logic bit value based on a 50% random assignment.

US Patent:

8407171, Mar 26, 2013

Filed:

Aug 10, 2010

Appl. No.:

12/853886

Inventors:

Mithun Das Gupta - Cupertino CA, US
Jing Xiao - Cupertino CA, US
Sanjeev Kumar - San Diego CA, US

Assignee:

Seiko Epson Corporation - Tokyo

International Classification:

G06F 17/00

US Classification:

706 45

Abstract:

Similarities between simplex projection with upper bounds and Lprojection are explored. Criteria for a-priori determination of sequence in which various constraints become active are derived, and this sequence is used to develop efficient algorithms for projecting a vector onto the L-ball while observing box constraints. Three projection methods are presented. The first projection method performs exact projection in O(n) worst case complexity, where n is the space dimension. Using a novel criteria for ordering constraints, the second projection method has a worst case complexity of O(n log n). The third projection method is a worst case linear time algorithm having O(n) complexity. The upper bounds defined for the projected entries guide the L-ball projection to more meaningful predictions.

US Patent:

8412757, Apr 2, 2013

Filed:

Dec 9, 2009

Appl. No.:

12/634291

Inventors:

Mithun Das Gupta - Cupertino CA, US
Jing Xiao - Cupertino CA, US

Assignee:

Seiko Epson Corporation - Tokyo

International Classification:

G06F 7/00

US Classification:

708207, 708200, 708446, 708514

Abstract:

Non-negative matrix factorization, NMF, is combined with identification of a maximum margin classifier by minimizing a cost function that contains a generative component and the discriminative component. The relative weighting between the generative component and the discriminative component are adjusting during subsequent iterations such that initially, when confidence is low, the generative model is favored. But as the iterations proceed, confidence increases and the weight of the discriminative component is steadily increased until it is of equal weight as the generative model. Preferably, the cost function to be minimized is:.

US Patent:

8498949, Jul 30, 2013

Filed:

Aug 11, 2010

Appl. No.:

12/854768

Inventors:

Seung-il Huh - Pittsburgh PA, US
Mithun Das Gupta - Cupertino CA, US
Jing Xiao - Cupertino CA, US

Assignee:

Seiko Epson Corporation - Tokyo

International Classification:

G06F 15/18
G06F 17/16

US Classification:

706 12, 708520

Abstract:

Supervised nonnegative matrix factorization (SNMF) generates a descriptive part-based representation of data, based on the concept of nonnegative matrix factorization (NMF) aided by the discriminative concept of graph embedding. An iterative procedure that optimizes suggested formulation based on Pareto optimization is presented. The present formulation removes any dependence on combined optimization schemes. Analytical and empirical evidence is presented to show that SNMF has advantages over popular subspace learning techniques as well as current state-of-the-art techniques.