RDD52 Guardbit requirements of JPEG 2000

What is a guard bit? And how do I know what is set in my encoded JPEG 200 image.

There are two freely available tools that will produce you the Guard Bit Information of a JPEG 2000 image.

OpenJPEG

Distributed under the open source project OpenJPEG. The included CLI utili opj_dump will produce a STDOUT dump.

Kakadu (kdu_show)

Kdu_winshow or kdu_macshow is a lightweight UI based JPEG 200 inspector tool.

Background

Understanding the SMPTE RDD 52 constraints on ISO 15444-1 2016.

NOTE: Remember that the 2016 standdard is explicitly referneced for D-Cinema applications.

“The number of guard bits to be specified in the QCD marker shall be a value of 1 for 2K content, and shall be a value of 2 for 4K content.”

QCD: Quantization Default describes the quantization default used for compressing all components. It is located in the main and first tile-part header of a given tile. This is applied to all components, each component being a tile-part of a given single tile, thus aligning with the RDD 52 constraint “the entire image shall be encoded as a single tile.” As such, there is a guard bit value for each component. This is summarized and presented differently in each of the three aforementioned tools which I will now detail.

Kakadu

1. Open you j2c image in kdu_show
2. Then go to Main Menu > File > Properties

3. In the Properties dialog, select Main Header. The right hand side of the UI will now populate with similar info pertaining to the j2c file.

Here kdu_show displays it differently, as all components = a part-tile of a main tile it displays the guard bit info (Qguard) for the main tile. Here it is set to ‘1’ which is correct for a 2K image.

Note: This would be set to ‘2’ if 4K.

OpenJPEG

Run the following command

opj_dump -i <path-to-j2c.j2k>

[INFO] Start to read j2k main header (0).
[INFO] Main header has been correctly decoded.
Image info {
	 x0=0, y0=0
	 x1=1998, y1=1080
	 numcomps=3
		 component 0 {
		 dx=1, dy=1
		 prec=12
		 sgnd=0
	}
		 component 1 {
		 dx=1, dy=1
		 prec=12
		 sgnd=0
	}
		 component 2 {
		 dx=1, dy=1
		 prec=12
		 sgnd=0
	}
}
Codestream info from main header: {
	 tx0=0, ty0=0
	 tdx=1998, tdy=1080
	 tw=1, th=1
	 default tile {
		 csty=0x1
		 prg=0x4
		 numlayers=1
		 mct=1
		 comp 0 {
			 csty=0x1
			 numresolutions=6
			 cblkw=2^5
			 cblkh=2^5
			 cblksty=0
			 qmfbid=0
			 preccintsize (w,h)=(7,7) (8,8) (8,8) (8,8) (8,8) (8,8)
			 qntsty=2
			 numgbits=1
			 stepsizes (m,e)=(1844,15) (1777,15) (1777,15) (1710,15) (1777,14) (1777,14) (1710,14) (1794,13) (1794,13) (1760,13) (1869,12) (1869,12) (1895,12) (3,10) (3,10) (68,10)
			 roishift=0
		 }
		 comp 1 {
			 csty=0x1
			 numresolutions=6
			 cblkw=2^5
			 cblkh=2^5
			 cblksty=0
			 qmfbid=0
			 preccintsize (w,h)=(7,7) (8,8) (8,8) (8,8) (8,8) (8,8)
			 qntsty=2
			 numgbits=1
			 stepsizes (m,e)=(1844,15) (1777,15) (1777,15) (1710,15) (1777,14) (1777,14) (1710,14) (1794,13) (1794,13) (1760,13) (1869,12) (1869,12) (1895,12) (3,10) (3,10) (68,10)
			 roishift=0
		 }
		 comp 2 {
			 csty=0x1
			 numresolutions=6
			 cblkw=2^5
			 cblkh=2^5
			 cblksty=0
			 qmfbid=0
			 preccintsize (w,h)=(7,7) (8,8) (8,8) (8,8) (8,8) (8,8)
			 qntsty=2
			 numgbits=1
			 stepsizes (m,e)=(1844,15) (1777,15) (1777,15) (1710,15) (1777,14) (1777,14) (1710,14) (1794,13) (1794,13) (1760,13) (1869,12) (1869,12) (1895,12) (3,10) (3,10) (68,10)
			 roishift=0
		 }
	 }
}
Codestream index from main header: {
	 Main header start position=0
	 Main header end position=171
	 Marker list: {
		 type=0xff4f, pos=0, len=2
		 type=0xff51, pos=2, len=49
		 type=0xff52, pos=51, len=20
		 type=0xff5c, pos=71, len=37
		 type=0xff55, pos=108, len=21
		 type=0xff64, pos=129, len=42
	 }
}

As before, opj_dump displays the same info differently, this time opting for a component by component layout. As you can see in the highlighted output, per component, “numgbits” is set to ‘1’.

Note: This would be set to 2 if 4K.