Large sharp barcode photos not correctly recognized about jjil HOT 7 CLOSED

The previous code didn't really do barcode detection. I've added a new 
DetectBarcode
class to the barcode library and used it in ReadBarJ. I believe you will find
performance is much better. ReadBarJ will show the area of the image it 
detected the
barcode and also outline the approximate barcode edges with green lines. I 
updated
the source tree and also uploaded a new JAR file for ReadBarJ.

I created a larger monochrome version of barcodeEAN.png (see attached 
barcode_clean.png image).  All pixels 
are 0x000000 or 0xFFFFFF.  Each bar is exactly 3, 6, 9, or 12 pixels wide.  The 
total width of the barcode is 
285 pixels, centered within a 640 x 480 image with 0xFFFFFF background.

The output from the latest J2SE build is "The barcode could not be read".  The 
debug images look correct (see 
attachments); and the cYLeft, cYRight, wSlopeLeft, and wSlopeRight values 
passed to 
BarcodeReaderEan13.Decode() look good.  szCode is estimated as 8269992866737, 
4084483768866, 
5003443366630, and 1906153945336.  Estimated offsets seem good, so it appears 
there may be a bug in 
estimating the left and right codes?  Or maybe my image is just too perfect 
whereas the correlations expect 
more randomness?

Reattaching the image I created, which didn't upload correctly the first time.

-BarcodeReaderEan13.java line 800-803:

// estimate position of the middle of the barcode
// this will be the position of the second 0 in the
// string "0101010"
int dMidOffset =

It seems from the code that the string should be "1010101" where dMidOffset 
estimates the position of the second 1 (the first black bar in the "01010" 
center guard pattern)?  At 
least this is what the code estimates (x=187 in my attached cropped.png).  Or 
is the intention to estimate the left edge of the center guard pattern?


-BarcodeReaderEan13.java line 813-816:

// estimate position of the right of the barcode.
// this will be the position of the second 0 in the
// string "1010000"
int dRightOffset = 

It appears the code actually estimates the position of the first 0 (x=328 in my 
attached cropped.png).  Is this intentional, or should it be matching the 
second 0 as suggested in the 
comment?

Narrowed the problem down to the following code from BarcodeReaderEan13.java 
(lines 1386-1393):

    private byte[] Stretch(byte[] bInput, int dLeft, int dRight, int dTargetWidth) {
        int dWidth = dRight - dLeft;
        byte[] bResult;
        if (dWidth % dTargetWidth == 0) {
            bResult = new byte[dTargetWidth];
            System.arraycopy(bInput, dLeft, bResult, 0, dTargetWidth);
            return bResult;
        }

in the case that (dWidth % dTargetWidth == 0), we should be creating a byte 
Array of width dWidth rather 
than dTarget width.  The current code truncates the array at the target width 
when the original data may have 
been a multiple of the target width.

For the sake of completeness, the fix is (BarcodeReaderEan13.java lines 
1390-1391):

            bResult = new byte[dWidth];
            System.arraycopy(bInput, dLeft, bResult, 0, dWidth);

Thanks, hansomli. I've made the change you suggested to the code and have given 
you
credit. The code now reads the 'original.png' image correctly. Please verify if 
you
have time.