A Java library to query pictures with SQL-like language.
Features :
- Select and manipulate pixels of pictures in your disk with SQL-like dialect (only BMP at this time).
- Access to each color channel r,b,g and pixel position : x,y, rank().
- Can query any number of pictures in the FROM clause, load only a region or subquery or build a colored rectangle.
- Math operators : *, /, +, -, %.
- Math functions: rand, pi, cos, sin, tan.
- Where clause with boolean operators : and / or.
- Lag and lead to get relative pixels from a position.
- Simple editor with execution of query in real time.
java -jar picsql.jar "select r,g,b from ./test.bmp" "./output.bmp"java -jar picsql.jar --guiFlip a channel on x, y or xy axis
Test pictures :
face.bmp
face2.bmp
select r, g, b
from./examples/face.bmp -- same image that original, we rebuild it.
select r, g, b
from (. / examples / face . bmp, 10, 50, 120, 120) -- (x,y,width,height) of a region
select r, g, b
from (./examples/face.bmp, 5, 5)
select r, g, b
from (100, 100, 255, 0, 0) -- (width,height, r, g, b)
select g, b, r
from./examples/face.bmp -- put green in red, blue in green and red in blue.
select (r * y)%255, x, b
from./examples/face.bmp
select x * 2, y + 10, x - 10
from./examples/face.bmp -- same size than face.bmp
select r, g, b
from./examples/face.bmp
where r > 120
select r, g, b
from./examples/face.bmp
where r > 20 and g < 200
select (r * sin(x))%255,
(g*tan(y))%255,
(cos(r))%255
from./examples/face.bmp
select lag(r, 10, 10), --lag select a red value at x-10, y-10
lead(g, 5, 5), --lead green value at +5,+5.
g
from./examples/face.bmp, 
select lag(r, sin(x), (r * g)%10),
lag(r, y%10, rank()%10),
lead(b, 10, 15)
from./examples/face.bmp
-- Lag, lead or other functions can be nested
select lag(r, lag(g, 5, 5)%5, 5),
lag(g, 10, lag(r, 5, 5)%5),
lag(r,
lag(g, 5, 15)%20,
lead(g, 15, 5)%10
)
from./examples/face.bmp
select f.r, -- reference the red of the first picture
f.g, -- reference the green of the first picture
f2.b -- reference to the blue of the second picture
from
./examples/face.bmp f, -- alias is necessary with more than one picture
./examples/face2.bmp f2
select (f.r + f2.b)%255,
lead(f.r, f2.b%10, 10),
f.g
from./examples/face.bmp f, ./examples/face2.bmp f2
select sub1.r,
sub2.g,
(sub2.b + sub1.r)%255
from
(select r from./examples/face.bmp where r > 60) sub1,
(select b, g, lag(r, 5, 5) from./examples/face2.bmp) sub2
A convolution mask is a 9 values array, that will be applied on each pixel of the picture. Each value in the array take a different value of a channel relative to the pixel. The result is the sum of the values divided by nine.
So if x is the pixel, and x1, x2, x3, x4, x5, x6, x7, x8 are the pixels around x, the convolution mask is :
x1 | x2 | x3
x4 | x | x5
x6 | x7 | x8
So value of red channel x.r is : (x1.r + x2.r + x3.r + x4.r + x.r + x5.r + x6.r + x7.r + x8.r) / 9
select r,
[-r,-b,-g,b*2,r,r,r,g*2,b],
[b,-b,b*2,r*r,b,g,b,-g,g]
from ./examples/face.bmp
Take a value of a channel and return a value between 0 and 255, with a step of 255 / step.
select
discr(r, 10),
discr(r, 5),
discr(g, 3)
from ./examples/face.bmp
select
flip(r, x),
flip(g, x),
flip(b, x)
from ./examples/face.bmp
Ugly example:
select
flip(r, x),
flip(g, xy),
flip(b, xy)
from ./examples/face.bmp
Select r, g or b color channel of the real butter color !
select
butter(r),
g,
butter(b)
from ./examples/face.bmp
Kalon vat !
select lag(r, 5, t%20), (g*t)%255, (t * 10) % 255
from (./examples/face.bmp, 50)Create a GIF by applying 50 times the query on the same picture, with increments "t" from 0 to 50.
Run query like that :
java -jar picsql.jar "select lag(r, 5, t%20), (g*t)%255, (t * 10) % 255 from (./examples/face.bmp, 50)" "test.gif"Result:
