how to convert int to uint8 in java

Question

I write an image filter in Matlab and I want to translate it to java in order to use my android application. How can I convert short or int to uint8 data format.

BTW:I know there is no type like uint8 in java.

Any help will be greatly appreciated.

Matlab:

for iX = 1:imageX
    for iY = 1:imageY
        x(1:3) = input(iX,iY,:);
        output(iX, iY, :) = uint8(p1.*x.^5 + p2.*x.^4 + p3.*x.^3 + p4.*x.^2 + p5.*x + p6);
    end
end

Java:

for (int iX = 0; iX < width; iX++) {
            for (int iY = 0; iY < height; iY++) {
                Arrays.fill(x, bitmap.getPixel(iX, iY));
                short total = 0;
                for (int i = 0; i < 3; i++) {
                    total +=p1[i] * Math.pow(x[i], 5) + p2[i]
                            * Math.pow(x[i], 4) + p3[i] * Math.pow(x[i], 3)
                            + p4[i] * Math.pow(x[i], 2) + p5[i] * x[i] + p6[i];
                }
                output.setPixel(iX, iY, total);
            }
        }

unfortunately, the value of total in java is different from matlab. — ibrahimyilmaz, Jan 18 '13 at 08:46
In matlab x is an representation of rgb color where - array of 3 elements where each element 1 byte? — Mikita Belahlazau, Jan 18 '13 at 08:48
Is `setPixel` last argument rgb color represented by java `int`? — Mikita Belahlazau, Jan 18 '13 at 08:49
i know byte is signed. that's why i said 8 bit int, not 8 bit uint. but you can work around that issue. signed-ness is a bit representation. what you would do is compute in short (or int or whatever) and then convert to byte for storing. — thang, Jan 18 '13 at 08:49
the reason why your java implementation outputs different values is not because of uint8 or anything. it's because you're using vector operations in matlab (i assume for rgb components), and you're summing everything up in java in a single short. they're different computations. — thang, Jan 18 '13 at 08:50
setPixel color parameter type is int but my math function should return uint8... — ibrahimyilmaz, Jan 18 '13 at 08:51
your math function should return 3 ints. you can then jam the 3 ints into 3-bytes of the int that setPixel takes. what is the type of output? — thang, Jan 18 '13 at 08:52
i think he's just trying to get it to work first... which apparently it currently isn't. — thang, Jan 18 '13 at 08:54
output is Bitmap and I know the performance issue but after accomplishing this task I will optimize... My main idea is to learn how to use uint8 like structure in java :) — ibrahimyilmaz, Jan 18 '13 at 08:56
@ibrahimyilmaz in matlab you don't use uint8. You are using array that later you convert to array of uint8. Do same in java. Use arrays. — Mikita Belahlazau, Jan 18 '13 at 08:58

shuangwhywhy · Answer 1 · 2013-01-18T09:37:03.037

3

Use a longer type in Java, for example, short<->uint8, int<->uint16, long<->uint32.

Or just use int for all those unsigned types under 32 bits and long for 32 bits:

When you operate on it, do & op in the end:

int i = 256 & 0xff;                             //uint8
int i = 65536 & 0xffff;                         //uint16
long i = (long) Math.pow(2, 32) & 0xffffffff;   //uint32;
// And So on

There is a "bit waste" as it always uses half of the type length. But in most cases, it won't be a serious problem.

edited Jan 18 '13 at 09:37

answered Jan 18 '13 at 09:18

shuangwhywhy

5,475
2
18
28

Caused by: java.lang.IllegalArgumentException: Expects arg[0] to be uint8 but int32 is provided it does not work when trying to do it in tensorflow. – JPerk Jun 12 '18 at 10:29

score 1 · Answer 2 · edited May 23 '17 at 12:07

1

This is an unfortunate limitation of java. This is what I would do.

Don't worry about the size, use an char or int
Write you own addition wrapper for byte using a bit mask to check the sign. This will be slow so I would go with #1. An alternative to this is to cast to a different data-type during the operation. Such as using an int before serializing the truncated version to a byte data-type.
See Is there a Java library for unsigned number type wrappers?

edited May 23 '17 at 12:07

Community

1
1

answered Jan 18 '13 at 08:50

Mikhail

7,749
11
62
136

1

Could you give an example please? – ibrahimyilmaz Jan 18 '13 at 08:52
1

http://stackoverflow.com/questions/8193031/is-there-a-java-library-for-unsigned-number-type-wrappers – Mikhail Jan 18 '13 at 09:05

Mikita Belahlazau · Accepted Answer · 2013-01-18T08:58:19.390

Don't create total. In matlab you handle each color component separately. Do same in java:

for (int iX = 0; iX < width; iX++) {
        for (int iY = 0; iY < height; iY++) {
            Arrays.fill(x, bitmap.getPixel(iX, iY));
            // Result color components.
            int[] res = new int[3];
            short total = 0;
            for (int i = 0; i < 3; i++) {
                res[i] = p1[i] * Math.pow(x[i], 5) + p2[i]
                        * Math.pow(x[i], 4) + p3[i] * Math.pow(x[i], 3)
                        + p4[i] * Math.pow(x[i], 2) + p5[i] * x[i] + p6[i];
                // Make sure final value is in range [0, 255]
                res[i] = Math.min(255, Math.max(0, res[i]));
            }
            output.setPixel(iX, iY, Color.rgb(res[0], res[1], res[2]);
        }
    }

Tuto · Answer 4 · 2013-01-18T08:58:23.113

0

Compiler will make an transformation for you:

int value = 115;
byte transformed = (byte) value;
// value == transformed => true

Warning: In case of transforming int to a byte you risk loosing 3 * 8 bits of information:

int value2 = 357;
byte transformed2 = (byte) value2;
// value2 == transformed2 => false

edited Jan 18 '13 at 08:58

answered Jan 18 '13 at 08:51

Tuto

92
6

score 0 · Answer 5 · answered Jan 18 '13 at 09:00

0

There is no unsigned data type in java so u will have to go with one out of them

int,char,short,byte.

These all are signed data type in java but keep in mind the type conversion rules while u assign the value of one data type value to another.

answered Jan 18 '13 at 09:00

Jatin Khurana

1,155
8
15

how to convert int to uint8 in java

5 Answers5