How to extract multiple mailing addresses from PDF file using pdftotext - regex

Question

I am using pdftotext in a bash script, trying to extract the names and addresses from PDF postage labels.
An example PDF file:
Delivered By:
1st Class
Postage on Account GB
First Last
HouseName
Street
Town
County
Postcode
Customer Reference: 12400 / 203 1
32224983765
RETURN TO: MyName,
DoorNumber, Street, Town,
City, Postcode, Country
121-0434 905 20200-000 6190 C228
Delivered By:
1st Class
Postage on Account GB
First Last
HouseNumber
Street
Town
Postcode
Customer Reference: 12401 / 200 1
32224286536
RETURN TO: MyName,
DoorNumber, Street, Town,
City, Postcode, Country
121-0434 905 20200-000 6190 C414
Please note:
The addresses do not have a fixed length ie. some consist of only 4 lines, and I have seen some with up to 6 lines.
The number of addresses in the PDF file will not be known in advance.
So far, I have just got:
pdftotext label.pdf - | grep -A10 "Postage on Account GB" | tail -n+3 | head -n -3
The - avoids creating a file. The grep -A10 outputs the first 10 lines from the match "Postage on Account GB". The tail -n+3 removes the match and next line. The head -n -3 removes the last 3 lines. That works fine when there's only one address in the file consisting of 6 lines, but I'm stuck when it comes to multiple addresses and with different lengths.
Put simply, I would like to extract the data from after the blank line after Postage on Account GB, until the line before the next blank line. Then format the output so that addresses are comma delimited and each on a new line, such as:
First Last, HouseName, Street, Town, County, Postcode
First Last, HouseNumber, Street, Town, Postcode

Answer 1

Updated Answer
In the light of your comments, I have updated my answer as follows:
pdftotext file.pdf - | perl -00 -wnl -e 'BEGIN{$a=$r=0} if($a){($add=$_)=~tr/\n/,/; $r=1; $a=0; next} if($r){printf "%s,%s\n",$_,$add;$r=0} $a=1 if m/Postage on Account/;'
One record is read each time through the loop - a record is separated by blank lines above and below because of -00. At the start, I set $a and $r flags to zero, meaning we are not looking at an address nor a reference. If we are looking at an address, I translate all newlines into commas and note that we are now looking for a reference. If we find a reference, we print it and the saved address and note that we are no longer looking at an address or a reference. If we find the string "Postage on Account", we note that we are now expecting an address to follow.
Sample Output
Customer Reference: 12400 / 203 1,First Last,HouseName,Street,Town,County,Postcode
Customer Reference: 12401 / 200 1,First Last,HouseNumber,Street,Town,Postcode
Original Answer
I think I'd go with Perl in paragraph mode:
pdftotext file - | perl -00 -wnl -e 'BEGIN{$p=1} if($p==1){tr/\n/,/;print;$p=0}; $p=1 if /Postage/'
The -00 sets Perl in paragraph mode treating each blank line delimited block as a paragraph. The BEGIN{...} sets the print flag ($p) so the first line gets printed. On subsequent paragraphs, when the print flag is set, the newlines get changed into spaces with tr and the paragraph gets printed and the flag reset. Finally, whenever we see the word Postage we set the print flag.

Answer 2

pdftotext filename.pdf - |sed -n '/Postage on Account GB/,/Customer Reference:/{/Postage on Account GB/!{/Customer Reference:/!p}}' |grep . |tr '\n' ',' |sed 's/,$//g' |sed "s/Postcode/&\n/g" |sed 's/^,//g'
First Last,HouseName,Street,Town,County,Postcode
First Last,HouseNumber,Street,Town,Postcode