I can also not see any misbehavior of Octave's conndef function, yet.

A typical usage of conndef would be the following:

pkg load image
C = conndef(2, "minimal")

The output is the connectivity matrix C, that can then be used by user image package functions.

@Yu Hongbo: Please read the help of this function (or of the Matlab equivalent function) and let us know in more details where you see a buggy behavior.

@Avinoam: If there is no further feedback of the OP, I think we could close this bug report then after some time.

First, the "input" below is a grayscale image matrix:

>> input

input =

   0.1000   0.2000   0.3000
   0.2000   0.3000   0.4000
   0.3000   0.4000   0.5000

>> pkg load image
>> conndef(input, 2)

error: conversion of 0.1 to unsigned int value failed

But, double data can be converted to unsigned int, like this:

>> uint8(0.1)

ans = 0

So, I cannot get where the true problem is.