Implement Hidden Markov Model in Matlab

I am new to HMM and although I understand the math (or I think I do), I do not fully understand how to implement it. I am given a series of about 50,000 lines coming in one at a time. Each line contains 100 elements (each element is a gray scale intensity level of 0-255).

The first 200 lines also contain a class label (one of 6 possible labels) of that line, so I can use them for training, but the following lines do not contain a label and I need to classify them as belonging to one of the 6 possible labels.

This is a sequence classification problem where I need to predict a single label to an entire input sequence. However, my difficulty is how to set up the problem, what is defined here as an “observation”? Also, is a “state” in HMM defined as an obvious feature of the data, meaning, does it represent a class label ?

I found many examples using a single binary input or output, but I still do not understand how I can apply HMM to my data. I would really appreciate it if you could help me set up the problem properly and which of the Matlab functions is suitable here.